MAPLE SDK ISSUE: CORE BUNDLE (maple-issue-2026-10-08)
=====================================================

Every text file from the maple-issue-2026-10-08 repository (bug report, standalone
reproduction project, SDK source excerpts and evidence), concatenated into one file.
This is the CORE bundle: report, repro code, SDK excerpts and evidence summaries/console logs.
The raw per-run evidence JSON/JSONL files cited by the report are only in the COMPLETE bundle.
There are no binaries or images in the repository. node_modules, build output and
.git are excluded. Run `bun install` inside repro/ to fetch dependencies.

FORMAT
------
Each file appears as:
  @@@@@@@@@@ FILE: <relative path> (<N> bytes)
  <exactly N bytes: the file contents, verbatim>
  @@@@@@@@@@ END FILE: <relative path>
(If a file does not end with a newline, one newline is inserted before the END
marker; the byte count in the header is the authoritative length.)

RECONSTRUCT THE ORIGINAL FOLDER, byte-for-byte (python3, run in an empty dir):
  python3 - THIS_FILE.txt <<'PY'
  import os, re, sys
  data = open(sys.argv[1], "rb").read()
  hdr = re.compile(rb"^@@@@@@@@@@ FILE: (.+?) \((\d+) bytes\)\n", re.M)
  pos = 0
  while (m := hdr.search(data, pos)):
      path, n = m.group(1).decode(), int(m.group(2))
      body = data[m.end():m.end() + n]
      os.makedirs(os.path.dirname(path) or ".", exist_ok=True)
      open(path, "wb").write(body)
      pos = m.end() + n
  PY

TABLE OF CONTENTS (44 files)
-----------------
  README.md  (2458 bytes)
  MAPLE_BUG_REPORT.txt  (212121 bytes)
  repro/README.md  (8750 bytes)
  repro/package.json  (379 bytes)
  repro/.gitignore  (153 bytes)
  repro/build.ts  (1913 bytes)
  repro/bun.lock  (4821 bytes)
  repro/client/app.ts  (9255 bytes)
  repro/client/probe.ts  (2686 bytes)
  repro/lib.ts  (7349 bytes)
  repro/public/index.html  (629 bytes)
  repro/public/probe.html  (293 bytes)
  repro/run-scenarios.ts  (8382 bytes)
  repro/scenarios/b-probe.ts  (9438 bytes)
  repro/scenarios/sdk.ts  (34500 bytes)
  repro/server.ts  (13999 bytes)
  repro/variants/sdk-0.10.0/bun.lock  (798 bytes)
  repro/variants/sdk-0.10.0/package.json  (415 bytes)
  sdk-excerpts/01-flush-core-cooldown-serialization.txt  (7677 bytes)
  sdk-excerpts/02-buffers-cap-restore.txt  (6299 bytes)
  sdk-excerpts/03-mapleflush-make-client.txt  (14768 bytes)
  sdk-excerpts/04-replay-transport-keepalive-budget.txt  (6957 bytes)
  sdk-excerpts/05-session-event-sink.txt  (4230 bytes)
  sdk-excerpts/06-session-lifecycle-unload.txt  (10642 bytes)
  sdk-excerpts/07-client-session-startup.txt  (9123 bytes)
  sdk-excerpts/08-replay-record-final-chunk.txt  (13459 bytes)
  sdk-excerpts/09-maple-layer-client.txt  (6922 bytes)
  sdk-excerpts/10-effect-otlp-exporter.txt  (8426 bytes)
  sdk-excerpts/11-readme.txt  (5172 bytes)
  sdk-excerpts/12-v0.10.0-what-changed-and-what-did-not.txt  (10614 bytes)
  evidence/headless-new/runner-console.txt  (29525 bytes)
  evidence/headless-new/summary-new-2026-10-07T23-50-02-862Z.txt  (13911 bytes)
  evidence/runner-console-A.txt  (5750 bytes)
  evidence/runner-console-F-rerun.txt  (900 bytes)
  evidence/runner-console-W.txt  (2528 bytes)
  evidence/runner-console-main.txt  (102949 bytes)
  evidence/sdk-0.10.0/runner-console-A.txt  (5863 bytes)
  evidence/sdk-0.10.0/runner-console-main.txt  (49901 bytes)
  evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-41-29-113Z.txt  (3149 bytes)
  evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-43-48-971Z.txt  (22139 bytes)
  evidence/summary-shell-2026-10-07T23-39-09-001Z.txt  (46826 bytes)
  evidence/summary-shell-2026-10-07T23-39-09-004Z.txt  (3061 bytes)
  evidence/summary-shell-2026-10-07T23-58-29-444Z.txt  (1271 bytes)
  evidence/webkit-firefox-launch.txt  (1867 bytes)

@@@@@@@@@@ FILE: README.md (2458 bytes)
# maple-issue-2026-10-08

Bug report and standalone reproduction for telemetry loss in the Maple browser SDK
(`@maple-dev/effect-sdk` 0.8.2, `@maple-dev/effect-sdk/client`, mainly `MapleFlush.make`).
Most findings still reproduce on 0.10.0.

- **[`MAPLE_BUG_REPORT.txt`](MAPLE_BUG_REPORT.txt):** the full report (~2,250 lines).
  - TL;DR table in section 1.
  - Per-bug root cause, evidence, reproduction steps and suggested fixes in section 4.
  - An index of every file in Appendix C.
- **[`repro/`](repro/):** self-contained Bun + Playwright project with a mock ingest server. See [`repro/README.md`](repro/README.md).
  ```sh
  cd repro && bun install && bun run scenarios      # all scenarios
  bun run scenarios A0 A                            # e.g. just the oversized-batch scenarios
  ```
- **[`sdk-excerpts/`](sdk-excerpts/):** verbatim SDK (and Effect) source excerpts with file names and line numbers.
- **[`evidence/`](evidence/):** raw output of every run cited in the report (one JSON per run, plus server request logs).

## Summary

| ID | Severity | Finding |
|---|---|---|
| BUG-1 | High | One flush whose keepalive body exceeds Chromium's 64 KiB in-flight budget (~163 spans) is rejected. It is then restored and cooled down for 60s, so the batch only grows and trace export stops for the rest of the page's life. |
| BUG-2 | Medium | traces, logs and metrics are POSTed concurrently with `keepalive` and share that budget. |
| BUG-3 | Medium | During the 60s cooldown the unload (`pagehide`) flush is skipped too. |
| BUG-4 | Medium | The unload flush is serialized behind any in-flight flush. With a slow ingest the tail is lost, and one slow POST stalls all signals. |
| BUG-5 | Medium | At unload, the OTLP posts and the session/replay posts compete for the same budget without shared accounting. |
| BUG-6 | Medium | The final replay chunk is never uploaded on unload (`MapleFlush.make` and `Maple.layer`). |
| BUG-7 | Low | "flush failed; cooldown 60s" / "POST failed" console errors at unload are mostly false alarms: the server received the requests and returned 200. |
| BUG-8 | Low | Non-keepalive session posts aborted by navigation are dropped, despite the "will retry on next chunk" message. |
| BUG-9 | Info | `Maple.layer` has no unload flush (documented). A small fix using Effect's own flusher is demonstrated. |

Environment: Chromium 143 (Playwright 1.57.0), Bun 1.4.2, Linux x86_64. WebKit and Firefox were not tested.
@@@@@@@@@@ END FILE: README.md

@@@@@@@@@@ FILE: MAPLE_BUG_REPORT.txt (212121 bytes)
MAPLE BUG REPORT
@maple-dev/effect-sdk 0.8.2, browser entry (@maple-dev/effect-sdk/client):
MapleFlush.make (and Maple.layer for comparison)

Telemetry gets lost or stops completely around fetch(keepalive), the 60 s
cooldown, the serialized flush and page unload. There are also misleading
console errors at unload, and Playwright reports requests as aborted when they
actually arrived.

Date of investigation: 2026-10-07
Reporter: a Maple customer, a React + Effect v4 single-page app (product details withheld)
Folder layout and index: see APPENDIX C at the end of this file.

How to read this report
-----------------------
* All empirical claims come from runs of the self-contained project in ./repro
  against a mock ingest in headless Chromium. The raw output of every run is in
  ./evidence (one JSON per run, plus the server's JSONL request log). Each finding
  cites the evidence files and scenario names.
* Each statement is labelled:
    [VERIFIED-EMPIRICAL]   reproduced in ./repro; evidence file(s) cited
    [VERIFIED-CODE]        follows directly from the quoted SDK source; not separately run
    [HYPOTHESIS]           plausible but not proven here
    [DISPROVEN]            we tested it and it was not true (kept so nobody re-investigates)
* SDK line numbers refer to the installed npm package
  (node_modules/@maple-dev/effect-sdk/dist/*.mjs, version 0.8.2). The verbatim
  excerpts are in ./sdk-excerpts (file names are given inline).
* "budget" = Chromium's per-document limit on the total body bytes of all
  keepalive fetches that are in flight at the same time (measured below: 65,536
  bytes).
* Throughout this report, "ingest" means the endpoint the SDK posts to. In our
  production setup that is a same-origin proxy (details in section 3). In the repro
  it is the mock server in ./repro/server.ts.


==============================================================================
1. TL;DR / SUMMARY OF FINDINGS
==============================================================================

Severity is our judgement of user impact: how much telemetry is lost and how
silently.

 ID     Sev.      Title                                                         Status
 ------ --------- ------------------------------------------------------------- --------------------
 BUG-1  HIGH      One oversized flush permanently stops trace (or log) export   VERIFIED-EMPIRICAL
                  for the rest of the page's life
 BUG-2  MEDIUM    traces + logs + metrics keepalive POSTs are fired             VERIFIED-EMPIRICAL
                  concurrently and share one 64 KiB budget, so one signal is
                  rejected although each fits on its own
 BUG-3  MEDIUM    The unload flush is skipped while a signal is in its 60 s     VERIFIED-EMPIRICAL
                  cooldown, so the failed batch plus everything since is lost
 BUG-4  MEDIUM    The unload flush is queued behind an in-flight flush           VERIFIED-EMPIRICAL
                  (makeSerializedFlush). With a slow ingest it only runs after
                  teardown has rejected the in-flight POST, which sets the
                  cooldown, so the unload flush is skipped and the tail is lost.
                  One slow POST also stalls ALL later exports (traces AND logs)
 BUG-5  MEDIUM    At unload, MapleFlush's OTLP POSTs and the session/replay      VERIFIED-EMPIRICAL
                  POSTs (meta "ended" row, final events) compete for the same
                  64 KiB budget with no shared accounting; whichever pagehide
                  listener runs first wins
 BUG-6  MEDIUM    The final replay chunk is never uploaded on unload (its        VERIFIED-EMPIRICAL
                  sequence number is consumed, then an async gzip runs, and the
                  document dies before the POST). Affects MapleFlush.make AND
                  Maple.layer
 BUG-7  LOW (but  False "failed" reports at every unload: keepalive requests     VERIFIED-EMPIRICAL
        expensive that the server answered 200 reject with "TypeError: Failed
        to triage) to fetch" in the unloading document. The SDK prints "traces/
                  logs/metrics flush failed; cooldown 60s", "... flush skipped
                  (cooldown 60000ms remaining)", and "[maple] session replay
                  metadata/events POST failed (dropping; will retry on next
                  chunk)". Playwright reports net::ERR_ABORTED
 BUG-8  LOW       Non-keepalive session posts (5 s events flush, "active"        VERIFIED-EMPIRICAL
                  metadata row, replay chunks) in flight during a navigation      (abort + message)
                  are aborted. The batch is dropped, and the console message      VERIFIED-CODE (drop)
                  claims "will retry on next chunk", which nothing does
 BUG-9  INFO /    Maple.layer has no unload flush (documented). A small fix is    VERIFIED-EMPIRICAL
        FEATURE   possible with Effect's own OtlpExporter.Flusher +
                  FetchHttpClient.RequestInit {keepalive:true}; we verified a
                  user-land version works
 MISC   LOW/INFO  Smaller items: section 6 (10k buffer cap keeps the oldest and   mixed, each labelled
                  drops the newest silently; keepaliveFor() counts UTF-16 code
                  units instead of bytes; transport never reads or cancels the
                  response body; Effect console logs are duplicated into
                  session events; "session replay" wording when replay is
                  disabled; the SDK sets a "user-agent" header; multi-argument
                  log bodies are sent as an OTLP arrayValue; ...)

What this means for our production observations (details in section 7):
 * The pair "[MapleClientSDK] traces flush failed; cooldown 60s: TypeError:
   Failed to fetch" followed immediately by "[MapleClientSDK] traces flush
   skipped (cooldown 60000ms remaining)" around a reload (also for "logs") is
   reproduced deterministically (scenario K: 3/3 runs replay on, 3/3 replay off).
   It is printed by the OLD document after pagehide, for requests the server
   answered 200. In that pattern it is a false alarm (BUG-7). It only appears when
   something was still buffered at pagehide, which explains why it looks rare and
   random. The same message can also mean real loss (BUG-1/2/5), and the SDK
   cannot tell the two cases apart.
 * "a span that ended ~5 s before a reload never arrived while its earlier sibling
   did": reproduced by BUG-4 (one slow ingest round-trip, 7 s here, is enough;
   3/3 runs). We cannot prove this was the production cause. It is our leading
   explanation because the alternatives would have printed console errors.
 * "[maple] session replay events/metadata POST failed (dropping; will retry on
   next chunk): TypeError: Failed to fetch" around navigations with both presets:
   mostly BUG-7 (data delivered; reproduced at every reload with pending session
   data). Real loss can come from BUG-8 (aborted non-keepalive posts) and BUG-5
   (budget).
 * "Playwright reports keepalive requests as requestfailed net::ERR_ABORTED although
   the data arrived": reproduced for every keepalive request still in flight when
   its document is torn down (G, K, C-control). In steady state with no navigation
   Playwright reports them as finished. This is an observability artifact of
   unload, not a delivery failure (section 5).

Newer SDK versions: 0.10.0 was published on 2026-10-03, after the version we run.
We read its source and ran the main scenarios against it (section 8). It adds a
shared keepalive budget for the session/replay POSTs and counts bytes correctly
there. The OTLP flush path (BUG-1, -2, -3, -4) and the final replay chunk (BUG-6)
look unchanged. See section 8 for the empirical results.


==============================================================================
2. ENVIRONMENT AND VERSIONS
==============================================================================

 Item                      Value
 ------------------------- ----------------------------------------------------------
 SDK under test            @maple-dev/effect-sdk 0.8.2 (npm), entry "@maple-dev/effect-sdk/client"
 Effect                    effect 4.0.0-rc.112 (the SDK's peer; also what our app uses)
 Also checked              @maple-dev/effect-sdk 0.10.0 + effect 4.0.0 (section 8)
 Browser                   Chromium 143.0.7499.4 (Playwright build 1200)
                           - default mode: chromium-headless-shell
                           - cross-check: full Chromium in "new headless" (channel "chromium")
                           UA: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like
                               Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36
 Playwright                1.57.0
 Bun                       1.4.2 (bundler: Bun.build target=browser, ESM, code splitting)
 OS                        Linux 6.18 x86_64 (Amazon Linux 2023 container), 8 vCPU, 16 GB
 Not tested                WebKit: Playwright's WebKit cannot launch on this host (missing
                           system libraries; see evidence/webkit-firefox-launch.txt).
                           Firefox: not installed. All browser results below are Chromium-only.
                           Safari/Firefox keepalive limits and unload behaviour differ and should
                           be checked separately.
 Network                   Everything on localhost: real network latency is NOT modelled
                           except where the mock server injects it (latencyMs rules). On a real
                           network, aborted requests are more likely to lose their body than
                           on localhost.

Exact environment record of each run: evidence/environment-*.json.


==============================================================================
3. HOW WE USE THE SDK (generic)
==============================================================================

* React + Effect v4 single-page app. All browser telemetry is produced by Effect
  (Effect.withSpan, Effect.log*, a few metrics) running on a ManagedRuntime that
  is provided the SDK's layer.
* Until recently: Maple.layer({ serviceName, environment, endpoint, ingestKey }).
  Now:
    MapleFlush.make({
      autoFlushInterval: 1000,
      ingestKey: "~",                 // placeholder; see below
      endpoint: "<same-origin>/otlp", // same-origin ingest proxy
      serviceName, environment,
    })
  Replay and session settings are left at their defaults (replay enabled,
  sampleRate 1, emitSessionMeta true).
* The endpoint is a same-origin proxy. The browser posts to <our origin>/otlp/v1/*
  and the proxy forwards to Maple ingest, adding the real ingest key on the server.
  The browser passes the placeholder ingestKey "~" only because the SDK switches
  itself off without a key. The proxy adds one extra network hop, so its latency
  varies; this matters for BUG-4.
* We test with Playwright-driven Chromium against deployed environments, which is
  where we made the observations in section 7.
* The repro page (repro/client/app.ts) mirrors this: same options, same
  placeholder key, endpoint = <page origin>/otlp/<run tag> (the tag only lets the
  mock server attribute requests to a run), and an optional cross-origin endpoint.


==============================================================================
4. FINDINGS IN DETAIL
==============================================================================

------------------------------------------------------------------------------
4.0 Ground truth: Chromium's fetch(keepalive) limit (scenario B-keepalive-probe)
------------------------------------------------------------------------------
We measured the browser behaviour first, independently of the SDK, because every
finding below depends on it. The probe page (repro/public/probe.html +
repro/client/probe.ts) has no Maple code on it. The mock sink can hold its
response for N ms so that requests stay in flight. Every sub-test starts from a
freshly loaded document. Evidence: evidence/B-keepalive-probe/probe-run0.json and
probe-run1.json (identical results in both runs).

 Sub-test                                                      Result (Chromium 143)
 ------------------------------------------------------------- ---------------------------------------
 B1 single keepalive POST, nothing else in flight              1024, 49152, 65535, 65536 B: ok
                                                               65537, 70000, 131072 B: REJECTED
                                                               (immediately, 2-7 ms, "TypeError:
                                                               Failed to fetch")
 B1b same with content-type: application/json                  65536 ok, 65537 REJECTED
 B2 concurrent, server holds responses 2 s:
    2 x 40000 B                                                1st ok, 2nd REJECTED
    2 x 30000 B                                                both ok
    30000 + 30000 + 10000 B                                    3rd REJECTED
    4 x 20000 B                                                1st-3rd ok, 4th REJECTED
 B3 hold 60000 B for 2 s; 10000 B at +500 ms; 10000 B after    held: ok; during: REJECTED; after: ok
 B3b 8 sequential 40000 B POSTs, each awaited:
    response body NOT read (what the SDK transport does)      "ok X X X X X X X"
    response body read (await res.text())                      all ok
    body not read, 50 ms gap between POSTs                      "ok ok X ok ok X ok ok" (run 0)
 B4 cross-origin with CORS preflight (application/json):
    1000 B                                                     ok (keepalive + preflight works)
    65537 B                                                    REJECTED
    2 x 40000 B concurrent                                     2nd REJECTED
 B5 200000 B WITHOUT keepalive                                 ok (no limit)
 B6 2-byte UTF-8 body ("é"): bytes / JS string length:
    60000 / 30000                                              ok
    65536 / 32768                                              ok
    80000 / 40000                                              REJECTED
    98304 / 49152                                              REJECTED
 B7 tab 1 holds 60000 B; then:
    tab 1, second 10000 B                                      REJECTED
    tab 2 (same context), 60000 B                              ok
    tab 3 (other context), 60000 B                             ok
 B8 keepalive POSTs fired from a pagehide handler, then reload:
    2 x 20000                                                  2/2 arrived
    2 x 40000                                                  1/2 arrived
    3 x 20000                                                  3/3 arrived
    4 x 20000                                                  3/4 arrived
    1 x 70000                                                  0/1 arrived
 B9 previous document still has 60000 B in flight; new          ok (the budget is per document)
    document sends 60000 B
 B10 50000 B keepalive in flight (held 4 s); pagehide fires a   only the 50000 B one arrived; the
     20000 B keepalive, then reload                            pagehide one was rejected

Conclusions [VERIFIED-EMPIRICAL]:
 * The limit is 65,536 bytes of UTF-8 body, summed over all keepalive requests of
   ONE document that are in flight at the same time. A request that would cross
   it is rejected synchronously with the generic "TypeError: Failed to fetch".
   There is no distinct error type or message, and the request never reaches the
   network.
 * A request keeps its share of the budget until its response body has finished
   loading, not just until fetch() resolves. If the body is never read, the bytes
   are released a few ms to tens of ms later (B3b).
 * The budget is per document: other tabs/contexts are unaffected, and a new
   document after a navigation starts with a fresh budget even while the previous
   document's keepalive requests are still in flight (B7, B9).
 * Keepalive requests in flight from earlier (for example a periodic flush to a
   slow ingest) reduce what is left for the unload flush (B10).
 * Cross-origin keepalive with preflight works in Chromium 143 (B4). The
   "preflight + keepalive unsupported" concern is [DISPROVEN] for current Chromium.
   We did not check other browsers.

------------------------------------------------------------------------------
4.1 BUG-1 (HIGH): one oversized flush permanently stops trace export for the
    rest of the page's life
------------------------------------------------------------------------------
Symptoms (exact console output, Chromium 143, from evidence/A-oversized-batch/*):
   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
       at window.fetch (http://localhost:4173/dist/chunk-<hash>.js:71:27)   <- replay engine's fetch wrapper
       at Object.post (http://localhost:4173/dist/app.js:...)                <- keepaliveTransport.post
       ...
 This repeats every ~60 s for as long as the tab stays open. In between, the
 auto-flush timer prints this every autoFlushInterval:
   [MapleClientSDK] traces flush skipped (cooldown NNNNNms remaining)
 No span reaches ingest again, including tiny spans created long after the burst.
 Other signals (logs) keep flowing, so dashboards look partly alive.

Root cause [VERIFIED-CODE], see sdk-excerpts/03-mapleflush-make-client.txt,
sdk-excerpts/01-flush-core-cooldown-serialization.txt and
sdk-excerpts/02-buffers-cap-restore.txt:

 (a) The client transport ALWAYS uses keepalive, whatever the body size
     (client-BYIixzTb.mjs L1989-1998):
       /** `fetch(keepalive)` transport — see file header for why not `sendBeacon`. */
       const keepaliveTransport = { post: async (url, headers, body) => {
           const res = await fetch(url, {
               method: "POST", headers, body: JSON.stringify(body), keepalive: true
           });
           if (!res.ok) throw new Error(`OTLP ${res.status} ${res.statusText}`);
       } };
     It is used for EVERY flush, including the periodic ones in steady state
     (L2070-2096), not only at unload.

 (b) flushSignal posts the WHOLE drained buffer as ONE request, with no chunking
     and no size check (version-B225m7RW.mjs L79-97):
       const batch = buffer.drain();
       if (batch.length === 0) return;
       const posted = await Effect.runPromise(Effect.result(Effect.tryPromise({
           try: () => transport.post(url, headers, body(batch)), ... })));
       if (Result.isFailure(posted)) {
           buffer.restore(batch);
           state.disabledUntil = Date.now() + COOLDOWN_MS;      // 6e4
           console.error(`${logPrefix} ${signal} flush failed; cooldown 60s:`, posted.failure);
       }

 (c) restore() puts the failed batch back IN FRONT of everything buffered since
     (version-B225m7RW.mjs L483-486, spans; L311-314, logs):
       restore: (items) => { if (disabled || items.length === 0) return;
                             buffer = [...items, ...buffer].slice(0, MAX_BUFFER); },
     The next attempt (60 s later) is therefore at least as large as the one that
     failed, plus 60 s of new spans. Once a batch exceeds 65,536 bytes it can never
     succeed again; the failing batch only grows.

 (d) At MAX_BUFFER = 1e4 (L422/L443) new spans are dropped silently in exportFn
     (`if (buffer.length >= MAX_BUFFER) return;`), and restore's slice(0, MAX)
     keeps the OLDEST items and drops the newest. Neither prints a warning or
     exposes a counter. [VERIFIED-CODE; not run to 10k]

 (e) The README states "keepalive carries the header and still survives unload
     for small bodies" (sdk-excerpts/11-readme.txt, README L277), but nothing
     keeps the bodies small.

Empirical evidence [VERIFIED-EMPIRICAL]:
 * Size per span and the exact threshold (scenario A0-threshold, autoFlushInterval
   false, one manual flush; evidence/A0-threshold/n*-run0.json). Body sizes are
   from Playwright's request.postDataBuffer().
     spans in one flush   traces POST body   result
     100                  40,252 B           delivered
     150                  60,052 B           delivered
     160                  64,012 B           delivered
     163                  65,200 B           delivered
     164                  65,596 B           REJECTED ("traces flush failed; cooldown 60s")
     165                  65,992 B           REJECTED
     170                  67,972 B           REJECTED
     200                  79,852 B           REJECTED
   That is ~396 B per span with no custom attributes (it already includes
   session.id and the resource), plus ~650 B of envelope. With the default
   autoFlushInterval 5000, a page that ends more than ~163 plain spans in 5 s, or
   fewer spans with larger attributes (error stack traces, URLs, payload
   summaries), hits this.
 * Long runs (scenario A-oversized-batch; each run in its own browser process;
   after the burst, one small "tick" span every 2 s and one log every 10 s;
   observed for 130 s):
     variant                                 runs  spans created  delivered  "flush failed" at (ms)
     burst 200 x 500 B, interval 5000         2/2   265            0          ~5400, ~70400
     burst 200 x 500 B, interval 1000         2/2   265            0          ~2240, ~63240, ~124240
     steady 40 spans/s for 10 s, int. 5000    2/2   438            152 (the   ~10500, ~75500
                                                                   first batch only;
                                                                   0/38 later ticks)
     steady 30 spans/s for 10 s, int. 5000    2/2   338            338        none (~150 spans
                                                                              ≈ 60 KB per batch,
                                                                              under the limit)
   In every reproducing run, logs (created alongside) were all delivered
   ("logs delivered 13" / "8"). Only the traces signal dies.
   Evidence: evidence/A-oversized-batch/*.json, evidence/runner-console-A.txt.
 * Flakiness: none. 100% deterministic for a given payload size.
   Per-variant verdict counts: burst200x500B-interval1000 -> [REPRODUCED (no span ever delivered): 2/2]; burst200x500B-interval5000(default) -> [REPRODUCED (no span ever delivered): 2/2]; steady30spansPerSec-10s-interval5000(default) -> [NOT REPRODUCED: 2/2]; steady40spansPerSec-10s-interval5000(default) -> [PARTIAL LOSS: 2/2]

Expected: telemetry keeps flowing after a burst. At worst the burst itself is
dropped or delayed, with a visible warning.
Actual: trace export from that tab is dead until the page is reloaded. Every
later span is lost. A console error appears once per minute (easy to miss in
production, where nobody reads the console).

Reproduction:
   cd repro && bun install && bun run build.ts
   bun run run-scenarios.ts A0 A          # A takes ~2.5 min (runs in parallel)
 or manually: bun run start, open http://localhost:4173/?tag=manual&variant=flush, then
   in DevTools: await T.spans(200, 500, "burst"); then T.span("later") every few seconds;
   watch repro/server-requests.jsonl (no /v1/traces POST ever arrives).

Suggested fixes:
 1. Split every flush into chunks below a per-request byte limit (for example
    <= 48 KiB including the envelope), computed from the serialized size in UTF-8
    bytes.
 2. Do not use keepalive for steady-state (visible-page) flushes. keepalive only
    matters while the document unloads. Use keepalive only from the
    pagehide/visibilitychange(hidden) path, and there respect a page-wide budget
    shared with the replay/session posts (see BUG-5).
 3. If a keepalive POST is rejected (TypeError before any response), retry
    immediately WITHOUT keepalive and/or with smaller chunks, instead of restoring
    and cooling down. TypeError from a keepalive request is almost always "too
    big", not "collector broken".
 4. Do not restore unboundedly. Cap what is restored, drop or split oversized
    batches, and keep the newest data rather than the oldest (or make this
    configurable).
 5. Make drops observable: an internal counter/metric (dropped spans/logs by
    reason), a one-time console warning when MAX_BUFFER is hit, and maybe an
    onDrop callback.

Workaround available to users today [VERIFIED-EMPIRICAL, scenario W]:
 * Install a window.fetch wrapper BEFORE MapleFlush.make() that turns
   keepalive:true into keepalive:false when the UTF-8 body is larger than ~60 KiB
   (repro/client/app.ts, "?shim=1"). Steady-state export then recovers; only
   oversized flushes during unload remain at risk. 
   without shim: [LOSS (burst 0/200, ticks 0/5): 2/2]
       run 0: LOSS (burst 0/200, ticks 0/5) | shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
       run 1: LOSS (burst 0/200, ticks 0/5) | shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
   with shim:    [all delivered: 2/2]
       run 0: all delivered | shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
       run 1: all delivered | shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
   Evidence: evidence/W-workaround-keepalive-downgrade-shim/*.json
 * Keep autoFlushInterval small (we use 1000) and keep span attributes small. This
   lowers the chance but cannot rule it out.

------------------------------------------------------------------------------
4.2 BUG-2 (MEDIUM): traces, logs and metrics are POSTed concurrently, so they
    compete for the same 64 KiB
------------------------------------------------------------------------------
Symptoms:
   [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
 (or "traces"/"metrics", depending on which request is issued last) when the
 SUM of the three bodies exceeds 65,536 B even though each fits on its own. That
 signal then enters BUG-1/BUG-3 territory (restore + 60 s cooldown).

Root cause [VERIFIED-CODE] (version-B225m7RW.mjs L132-173, runFlush):
     await Promise.all([ flushSignal({... traces ...}), flushSignal({... logs ...}),
                         flushSignal({... metrics ...}) ]);
 Each flushSignal issues its own keepalive fetch, so all three are in flight
 together. There is no accounting across them, and none with the replay engine's
 keepalive posts.

Empirical evidence [VERIFIED-EMPIRICAL] (scenario B-sdk-concurrent-signals;
40 spans x 500 B attribute + 30 log lines x 800 B + 1 counter update; replay on):
     trigger                     runs   request bodies issued (Playwright)          result
     manual T.flush()            3/3    traces 38,472 B, logs 32,135 B, metrics 931  spans 40/40, logs 0/30
                                                                                     "logs flush failed; cooldown 60s"
     reload (pagehide flush)     3/3    same + session meta 2,524 B, events          spans 40/40, logs 0/30,
                                        40,248 B                                     session events batch lost,
                                                                                     metrics delivered
 Evidence: evidence/B-sdk-concurrent-signals/manual-flush-run{0,1,2}.json,
 reload-run{0,1,2}.json. Flakiness: none (6/6). In the reload case the console
 shows "metrics flush failed" and "traces flush failed" although both arrived (the
 BUG-7 artifact), while the lost logs and events print nothing reliable. During
 unload the console is not a usable signal.

Expected: three small requests all succeed.
Actual: whichever request is issued last and crosses 65,536 B is rejected. Here
that is the logs POST.

Suggested fixes: send the signals sequentially when using keepalive, or combine
them under one budget/byte accounting (and share it with the replay posts, see
BUG-5); apply BUG-1 fixes 1-3.

------------------------------------------------------------------------------
4.3 BUG-3 (MEDIUM): the unload flush is skipped during the 60 s cooldown
------------------------------------------------------------------------------
Symptoms (exact, from evidence/C-unload-during-cooldown/inject503-then-reload-run0.json):
   [MapleClientSDK] traces flush failed; cooldown 60s: Error: OTLP 503 Service Unavailable
   [MapleClientSDK] traces flush skipped (cooldown 59xxxms remaining)    <- auto-flush ticks
   ... reload ...
   [MapleClientSDK] traces flush skipped (cooldown 5xxxxms remaining)    <- pagehide flush
   [MapleClientSDK] traces flush skipped (cooldown 5xxxxms remaining)    <- visibilitychange flush

Root cause [VERIFIED-CODE] (version-B225m7RW.mjs L79-84):
     if (state.disabledUntil && Date.now() < state.disabledUntil) {
         console.warn(`${logPrefix} ${signal} flush skipped (cooldown ...ms remaining)`);
         return;
     }
 The pagehide/visibilitychange handlers (client-BYIixzTb.mjs L2141-2151) call the
 same flush() as the timer, so the cooldown applies to the LAST chance to send.
 Everything buffered goes down with the page: the restored failed batch plus
 every span created during the cooldown.

Empirical evidence [VERIFIED-EMPIRICAL] (autoFlushInterval 1000; the mock returns
503 for exactly ONE traces POST; span "C-first" is created, its POST gets the 503,
then 1.5 s later "C-before-unload" is created and the page is unloaded 300 ms
after that):
     unload method                          runs   C-first      C-before-unload
     page.reload()                          3/3    LOST         LOST
     navigate to another same-origin page   3/3    LOST         LOST
     page.close({runBeforeUnload:true})     3/3    LOST         LOST
     control: no injected failure, reload   3/3    delivered    delivered
 One transient 5xx (or one network blip, or one BUG-1/BUG-2 rejection) within the
 last 60 s of a page's life loses that page's tail. With autoFlushInterval 1000,
 a single failure skips 60 flushes.
 Evidence: evidence/C-unload-during-cooldown/*.json. Flakiness: none (9/9).

Expected: the unload flush always tries to send, because there is no later
chance.
Actual: it is skipped; the data is lost and only a console.warn records it.

Suggested fixes: never apply the cooldown to the unload path (pagehide /
visibilitychange hidden / dispose). Better, replace the fixed 60 s cooldown with
exponential backoff that resets on success, and distinguish "collector rejected
(4xx/5xx)" from "keepalive rejected (TypeError, no response)".

------------------------------------------------------------------------------
4.4 BUG-4 (MEDIUM): the unload flush is serialized behind an in-flight flush;
    one slow request stalls all signals
------------------------------------------------------------------------------
Root cause [VERIFIED-CODE] (version-B225m7RW.mjs L115-123):
     const makeSerializedFlush = (run) => {
         let tail = Promise.resolve();
         return (...args) => {
             const next = tail.then(() => run(...args), () => run(...args));
             tail = next.catch(() => void 0);
             return next;
         };
     };
 All flush() calls (timer, pagehide, visibilitychange, manual) go through this
 chain (client-BYIixzTb.mjs L2070). If a flush is waiting on a slow POST when
 pagehide fires, the pagehide flush can only start after that POST settles. The
 keepalive transport has no timeout either, so one hung request stalls the chain
 indefinitely, and setInterval keeps adding a closure per tick.

 What actually happens in Chromium [VERIFIED-EMPIRICAL] (page journal of all 12
 D runs on 0.8.2 and all 8 on 0.10.0):
   1. The slow POST is still in flight when pagehide fires. The pagehide and
      visibilitychange flushes, plus every timer flush queued while waiting, sit
      in the chain.
   2. During teardown Chromium rejects the in-flight POST's promise with
      "TypeError: Failed to fetch". The request itself continues: the server
      later answers 200 and the early span IS delivered. flushSignal treats the
      rejection as a failure: restore + 60 s cooldown +
      "[MapleClientSDK] traces flush failed; cooldown 60s".
   3. The chain advances, and every queued flush, including the pagehide one,
      runs within the same millisecond and is SKIPPED:
      "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)"
      (twice in the 4 s variants, 7 times in the 7 s variant).
   So the tail is lost through the combination of serialization (BUG-4), the
   teardown false failure (BUG-7) and the cooldown at unload (BUG-3).
   [HYPOTHESIS] Removing only the cooldown on that path might let the queued
   flushes send in time, since JS was evidently still running. Removing the
   serialization for the unload flush is the robust fix.
 runFlush awaits traces+logs+metrics together (Promise.all), so a slow traces
 POST also blocks the logs and metrics exports.

Empirical evidence [VERIFIED-EMPIRICAL]:
 * Scenario D-pagehide-behind-inflight-flush. The mock holds only the FIRST
   traces POST for N ms (models one slow round-trip through a proxy). The test
   waits until that POST has reached the server, creates "D-tail", and reloads
   150 ms later:
     (a) first POST held 4000 ms, autoFlushInterval 1000, D-tail ends 150 ms before reload
         [REPRODUCED (tail span lost): 3/3]
       run 0: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5125ms); D-tail delivered=false; reload at t=1401ms
       run 1: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5107ms); D-tail delivered=false; reload at t=1382ms
       run 2: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5108ms); D-tail delivered=false; reload at t=1375ms
     (b) first POST held 4000 ms, autoFlushInterval 5000, D-tail ends 150 ms before reload
         [REPRODUCED (tail span lost): 3/3]
       run 0: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9124ms); D-tail delivered=false; reload at t=5383ms
       run 1: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9115ms); D-tail delivered=false; reload at t=5367ms
       run 2: REPRODUCED (tail span lost) | first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9114ms); D-tail delivered=false; reload at t=5373ms
     (c) first POST held 7000 ms, autoFlushInterval 1000, D-tail ends 5000 ms (!) before reload
         [REPRODUCED (tail span lost): 3/3]
       run 0: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
       run 1: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
       run 2: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms
     (d) control: no latency, autoFlushInterval 1000
         [CONTROL OK: 3/3]
       run 0: CONTROL OK | first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1405ms); reload at t=1393ms
       run 1: CONTROL OK | first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1412ms); reload at t=1406ms
       run 2: CONTROL OK | first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1123ms); D-tail delivered=true (received t=1401ms); reload at t=1394ms
   In variant (c) the span ended 5 s before the reload and was still
   lost, because the 5 queued timer flushes and the pagehide flush were all
   waiting behind the one slow POST. This matches our production observation
   (section 7, obs. 1b).
 * Scenario D-hang-head-of-line-blocking (first traces POST held 15 s; one span
   and one Effect log created every second; no navigation): 
[REPRODUCED (exports stalled behind one slow request): 2/2]
       run 0: REPRODUCED (exports stalled behind one slow request) | first traces POST held 15000 ms (from t=1153ms): span export delay max 14990 ms / min 909 ms; log export delay max 14991 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17143:1sp/1157B 18143:1sp/1157B 19144:1sp/1157B]
       run 1: REPRODUCED (exports stalled behind one slow request) | first traces POST held 15000 ms (from t=1137ms): span export delay max 14977 ms / min 905 ms; log export delay max 14977 ms (logs endpoint itself is fast); traces POSTs at server t=[1116:1sp/1160B 16120:15sp/6677B 17115:1sp/1157B 18115:1sp/1157B 19115:1sp/1157B]
 * Probabilistic version (scenario Dr-random-reload-latency: a span every 100 ms,
   reload at a random time 2-6 s in, every traces POST delayed by the given
   latency, 10 runs each):
        latency150-interval1000            runs with loss 1/10; spans lost 1/426; runs where a traces POST was in flight at reload: 1/10
         latency 150 ms, interval 1000: created 40, lost 1 (created 100 ms before reload); trace flush in flight at reload: true
     latency600-interval1000            runs with loss 3/10; spans lost 11/423; runs where a traces POST was in flight at reload: 5/10
         latency 600 ms, interval 1000: created 54, lost 5 (created 513,411,305,202,100 ms before reload); trace flush in flight at reload: true
         latency 600 ms, interval 1000: created 22, lost 2 (created 202,100 ms before reload); trace flush in flight at reload: true
         latency 600 ms, interval 1000: created 34, lost 4 (created 406,304,202,100 ms before reload); trace flush in flight at reload: true
     latency600-interval5000            runs with loss 0/10; spans lost 0/370; runs where a traces POST was in flight at reload: 2/10
     control-latency0-interval1000      runs with loss 0/10; spans lost 0/415; runs where a traces POST was in flight at reload: 0/10
   Loss happened only in runs where a traces POST was in flight at the reload
   AND spans had ended after that POST started. The two latency600-interval5000
   runs flagged "in flight" lost nothing because the in-flight POST already
   carried every span (the reload came right after the 5 s tick; see their
   serverRecords). With ~150 ms latency and autoFlushInterval 1000 we saw 1 of 10
   reloads lose the last ~100 ms of spans. With 600 ms, 3 of 10 lost up to the
   last ~0.5 s. Production latency through a proxy is often in this range.
 Evidence: evidence/D-pagehide-behind-inflight-flush/*.json,
 evidence/D-hang-head-of-line-blocking/*.json, evidence/Dr-random-reload-latency/*.json.

Expected: the unload flush sends whatever is buffered at pagehide, whatever any
earlier request is doing. An unresponsive request does not block other signals.
Actual: the unload flush waits for the earlier request and the tail is lost; a
slow traces POST also delays logs.

Suggested fixes:
 * The unload flush should bypass the serialization. It can drain and send
   immediately, concurrently with any in-flight flush. Draining is synchronous,
   so there is no double-send risk. To keep the unload flush from being rejected,
   reserve budget for it (see BUG-5).
 * Add a timeout (AbortSignal.timeout) to steady-state (non-keepalive) POSTs.
   Restore and retry on timeout.
 * Do not let one signal's slow POST hold back the others (flush signals
   independently).
 * Collapse queued timer flushes: if a flush is already queued, a new tick
   should not enqueue another one. (0.10.0 adds coalesceSameArguments, but only
   for Workers; section 8.)

Workaround available today: make the ingest proxy acknowledge immediately (for
example 202 before forwarding upstream) so client round-trips stay short. Not
tested in this repro, but it follows from the mechanism.

------------------------------------------------------------------------------
4.5 BUG-5 (MEDIUM): MapleFlush OTLP posts and session/replay unload posts share
    one budget with no shared accounting
------------------------------------------------------------------------------
On pagehide, two independent parts of the SDK fire keepalive requests:
  (1) MapleFlush.make's own pagehide/visibilitychange handlers -> flush() ->
      traces/logs/metrics, keepalive:true, any size (client-BYIixzTb.mjs L2141-2151).
  (2) The session lifecycle's pagehide handler -> endRun({flush:true, keepalive:true})
      -> metadata "ended" row + final session-events batch (+ the final replay
      chunk, see BUG-6). Each uses keepalive only if keepaliveFor(true, body.length)
      (client-BYIixzTb.mjs L603-607, L1531-1573, L1599-1610; replay-DBMTP4cB.mjs
      L10762-10771).
The SDK's own comment states the constraint (client-BYIixzTb.mjs L594-603):
     * The Fetch spec caps the *combined* inflight keepalive body at 64 KiB, and on
     * the way out we issue up to three of these at once (metadata row, final events
     * batch, last replay chunk). ...
     const MAX_KEEPALIVE_BYTES = 49152;
     function keepaliveFor(requested, bytes) { return requested && bytes <= MAX_KEEPALIVE_BYTES; }
 but the check is per request: three 48 KiB bodies pass it and total 144 KiB.
 It also ignores the OTLP posts in (1) and anything still in flight. It also
 measures body.length (UTF-16 code units) where Chromium counts UTF-8 bytes (B6).
 The order in which the listeners fire decides who gets the budget:
   * replay enabled (default): the replay engine loads through a dynamic import,
     so its pagehide listener is registered AFTER MapleFlush's. OTLP goes first,
     and the session rows lose.
   * replay disabled / not sampled: startClientSession() starts the metadata-only
     lifecycle synchronously inside MapleFlush.make(), BEFORE MapleFlush registers
     its own pagehide listener. The session rows go first, and the TRACES lose.
 (Verified by request order in evidence: E2 vs E3, and K replay-on vs replay-off.)

Empirical evidence [VERIFIED-EMPIRICAL] (scenario E-combined-unload-budget; the
page loads, the first replay chunk is shipped, then a payload is created and the
page is reloaded ~300 ms later):
      E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
       run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
       run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
       run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
   E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 3/3]
       run 0: LOST: meta(ended), final sessionEvents | spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
       run 1: LOST: meta(ended), final sessionEvents | spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
       run 2: LOST: meta(ended), final sessionEvents | spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
   E3-replay-disabled-spans150(~60kB)  [LOST: traces: 3/3]
       run 0: LOST: traces | spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
       run 1: LOST: traces | spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
       run 2: LOST: traces | spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
   E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
       run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
       run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
       run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
   E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
       run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
       run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
       run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) | spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
 Also scenario B-sdk-concurrent-signals/reload (4.2): sessionEvents 40,248 B and
 logs 32,135 B were rejected in favour of traces 38,472 B + metrics + meta (3/3).
 Evidence: evidence/E-combined-unload-budget/*.json.

Expected: the SDK's unload writes are coordinated so the important ones (spans,
"ended" row) fit, and anything that does not fit falls back to a normal request.
Actual: the first to fire wins; the others are rejected and silently lost (the
console says "failed" both for these and for successfully delivered requests; see
BUG-7).

Suggested fixes: one page-wide keepalive byte budget (on globalThis, shared by
both SDK copies) that every keepalive request reserves from and releases when the
response body ends (0.10.0 does this for session posts only; the OTLP transport
must use it too). Measure bytes with TextEncoder. At unload, prioritise
(spans > ended row > events > replay chunk) and send what does not fit without
keepalive.

------------------------------------------------------------------------------
4.6 BUG-6 (MEDIUM): the final replay chunk is never uploaded on unload
------------------------------------------------------------------------------
Root cause [VERIFIED-CODE] (replay-DBMTP4cB.mjs L10627-10653):
     const flush = async (keepalive = false) => {
         if (stopped || exhausted || parts.length === 0) return;
         const body = `[${parts.join(",")}]`;
         ...
         const seq = nextChunkSeq();      // persisted to sessionStorage immediately
         resetBuffer();
         let gzipped;
         try { gzipped = await gzip(new TextEncoder().encode(body)); }   // CompressionStream: async
         ...
         if (await postSessionBlob(config, {...}, gzipped, keepalive) === "exhausted" ...
 On pagehide (onSuspend, L10762-10771) this runs with keepalive=true, but the
 POST is only issued after the CompressionStream round-trip finishes. The
 document is torn down before that, so the request is never created. The chunk
 sequence number has already been consumed and persisted, so the backend sees a
 permanent gap. The last seconds before every navigation/close (often the most
 interesting part of a replay) are never recorded.

Empirical evidence [VERIFIED-EMPIRICAL] (scenario E: after the first chunk (seq
0) has been uploaded, mutate the DOM, reload, wait for the new page's first chunk):
      E1-spans45-plus-60kB-DOM-mutation: run 0: replay chunkSeqs at server [0,2]; run 1: replay chunkSeqs at server [0,2]; run 2: replay chunkSeqs at server [0,2]
   E2-spans150(~60kB)-plus-30-console-lines: run 0: replay chunkSeqs at server [0,1]; run 1: replay chunkSeqs at server [0,1]; run 2: replay chunkSeqs at server [0,1]
   E4-no-spans-small-DOM-mutation: run 0: replay chunkSeqs at server [0,2]; run 1: replay chunkSeqs at server [0,2]; run 2: replay chunkSeqs at server [0,2]
   E5-Maple.layer-small-DOM-mutation: run 0: replay chunkSeqs at server [0,2]; run 1: replay chunkSeqs at server [0,2]; run 2: replay chunkSeqs at server [0,2]
 Playwright saw no request at all for the old page's final blob, so it is not a
 budget rejection: the request is never created.

Expected: the final chunk is sent (keepalive if it fits, otherwise a normal POST).
Actual: it is silently never sent, and the chunk sequence has a hole.

Suggested fixes: compress incrementally while recording, so a ready-to-send gzip
buffer exists at pagehide; or on unload send the chunk uncompressed (or with a
synchronous compressor) inside the pagehide task; or keep the last chunk in
sessionStorage/IndexedDB and upload it from the next page load of the same
session. Claim the sequence number only when the POST is actually issued.

------------------------------------------------------------------------------
4.7 BUG-7 (LOW severity, HIGH triage cost): false failure reports at every unload
------------------------------------------------------------------------------
Symptoms (exact, scenario K: span + Effect log, reload 150 ms later,
autoFlushInterval 1000; evidence/K-reload-150ms-after-action/*):
   [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
   [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
   [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
   [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
   [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)
 and Playwright: requestfailed net::ERR_ABORTED for /v1/traces, /v1/logs,
 /v1/sessionReplays/meta, /v1/sessionEvents.
 Meanwhile the mock ingest received all four requests and answered 200
 (server timestamps 2-10 ms BEFORE the browser reported the failure).

Which document and which flush? [VERIFIED-EMPIRICAL] The page mirrors every
[MapleClientSDK]/[maple] console message, synchronously, into a localStorage
journal tagged with a per-document id and a flag "pagehide already fired". Every
one of these messages carries the OLD document's id, afterPagehide=true and
visibilityState "hidden" (example: evidence/K-reload-150ms-after-action/
replay-on(default)-run0.json -> pageJournal). The order is:
     pagehide -> visibilitychange:hidden -> "... flush failed" -> "... flush skipped"
     -> new document script-start.
So:
  * "flush failed" belongs to the pagehide flush. Its keepalive POSTs were issued,
    delivered and answered 200, but the fetch() promises of the unloading
    document reject with TypeError: Failed to fetch, because the document can no
    longer receive the responses. The keepalive request itself completes in the
    browser process.
  * "flush skipped (cooldown 60000ms remaining)" is the visibilitychange(hidden)
    flush, which ran within the same millisecond and found the cooldown that the
    previous line had just set. The value 60000/59999 always identifies this case.
  * Nothing fails in the NEW document.
  * The messages only appear if something was buffered at pagehide (otherwise
    flushSignal returns before fetch), which explains why they look intermittent.
 Results (scenario K):
   replay on (default): 
[span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 3/3]
       run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] | reload at t=1795; requests issued around the reload (Playwright): traces(1165B)@1807 logs(1014B)@1807 sessionReplays/meta(1164B)@1807 sessionEvents(1021B)@1807 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1801 logs@1802 sessionReplays/meta@1802 sessionEvents@1804 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1808 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1808 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1808 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1808
       run 1: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] | reload at t=1810; requests issued around the reload (Playwright): traces(1165B)@1824 logs(1014B)@1824 sessionReplays/meta(1164B)@1824 sessionEvents(1021B)@1824 sessionReplays/meta(1060B)@1891; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1817 sessionReplays/meta@1818 logs@1818 sessionEvents@1820 sessionReplays/meta@1890; console: [MapleClientSDK] logs flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1825 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1825 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1825 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1825
       run 2: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] | reload at t=1803; requests issued around the reload (Playwright): traces(1165B)@1815 logs(1014B)@1816 sessionReplays/meta(1164B)@1816 sessionEvents(1021B)@1816 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1809 logs@1810 sessionReplays/meta@1813 sessionEvents@1814 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1816 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1816 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1816 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1816
   replay off:          
[span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 3/3]
       run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] | reload at t=1770; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1781 sessionEvents(499B)@1781 traces(1165B)@1782 logs(1014B)@1782 sessionReplays/meta(1070B)@1814; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1777 traces@1778 sessionEvents@1779 logs@1781 sessionReplays/meta@1811; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1787 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1787 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1788 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1782
       run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] | reload at t=1771; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1783 sessionEvents(499B)@1783 traces(1165B)@1784 logs(1014B)@1784 sessionReplays/meta(1070B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1778 sessionEvents@1779 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1785 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1785 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1785 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
       run 2: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] | reload at t=1769; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1780 sessionEvents(499B)@1780 traces(1165B)@1781 logs(1014B)@1781 sessionReplays/meta(1070B)@1813; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1776 traces@1777 sessionEvents@1778 logs@1780 sessionReplays/meta@1810; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1782 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1782 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1782 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1781
 The same pattern appears in every control run that reloads with pending data
 (C-control 3/3, G-with-final-reload, E-*).

Why it matters: the SDK cannot tell "rejected by the budget, data lost" (BUG-1/2/5)
from "delivered, document gone" (this item). Both produce the same TypeError and
the same console line. Users chasing real loss get flooded with false alarms at
every navigation; users seeing false alarms learn to ignore the line that also
signals real loss. The "dropping; will retry on next chunk" wording is also wrong
(BUG-8).

Suggested fixes: record the phase (set a flag on pagehide). For requests issued
during unload, do not log, do not cool down, do not restore (the page is going
away). Optionally log once at debug level: "N bytes handed to the browser at
unload". Classify errors: a TypeError raised synchronously at fetch() time is a
budget rejection (real loss); a TypeError after unload is not actionable.

------------------------------------------------------------------------------
4.8 BUG-8 (LOW): non-keepalive session posts are aborted by navigation and
    dropped; the console message promises a retry that never happens
------------------------------------------------------------------------------
Root cause [VERIFIED-CODE]:
 * The session event sink flushes every 5 s with keepalive=false
   (client-BYIixzTb.mjs L1165-1171, L1198). The lifecycle posts its "active" row
   at start and its heartbeat with keepalive=false (L1512-1530). Replay chunks are
   flushed every 5 s or at 100 kB with keepalive=false (replay-DBMTP4cB.mjs
   L10681, L10689-10696). These are all aborted if a navigation happens while
   they are in flight.
 * On failure the batch has already been taken out of the buffer
   (`const batch = buffer; buffer = []`) and is not restored. warnDropped
   (L588-593) then prints "(dropping; will retry on next chunk)". Nothing retries
   the dropped batch; only future data gets sent. The message is also throttled
   to once per 30 s per kind, so repeated losses are under-reported.

Empirical evidence [VERIFIED-EMPIRICAL] (scenarios F and Fr; evidence/F-*/, evidence/Fr-*/).
 F "periodic-events": a console line is created; the mock holds the 5 s periodic
 (non-keepalive) sessionEvents POST for 3 s; the page reloads while that POST is in
 flight. F "active-meta": the mock holds every sessionReplays/meta POST for 3 s;
 the page reloads ~300 ms after load while the new page's "active" row is in
 flight. Fr: 300 ms latency on every ingest path, 8 reloads at random 0.5-7 s
 intervals. Counts come from the per-document page journal. Playwright's console
 capture showed the same lines, but during unload it is not guaranteed to.
      F-replay-posts-aborted-by-navigation / flush-active-meta-POST-in-flight run 0:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 2, 'replay metadata POST (after pagehide)': 2}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 4, 'sessionEvents[net::ERR_ABORTED]': 2}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 5, ('sessionEvents', 200): 2}
       evidence: evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run0.json
   F-replay-posts-aborted-by-navigation / flush-active-meta-POST-in-flight run 1:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 2, 'replay metadata POST (after pagehide)': 2}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 4, 'sessionEvents[net::ERR_ABORTED]': 2}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 5, ('sessionEvents', 200): 2}
       evidence: evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run1.json
   F-replay-posts-aborted-by-navigation / flush-periodic-events-POST-in-flight run 0:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 1, 'replay events POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 1, 'sessionEvents[net::ERR_ABORTED]': 1}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1, ('sessionReplays/blob', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run0.json
   F-replay-posts-aborted-by-navigation / flush-periodic-events-POST-in-flight run 1:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 1, 'replay events POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 1, 'sessionEvents[net::ERR_ABORTED]': 1}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1, ('sessionReplays/blob', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run1.json
   F-replay-posts-aborted-by-navigation / layer-active-meta-POST-in-flight run 0:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 1, 'replay metadata POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionEvents[net::ERR_ABORTED]': 1, 'sessionReplays/meta[net::ERR_ABORTED]': 2}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run0.json
   F-replay-posts-aborted-by-navigation / layer-active-meta-POST-in-flight run 1:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 1, 'replay metadata POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionEvents[net::ERR_ABORTED]': 1, 'sessionReplays/meta[net::ERR_ABORTED]': 2}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run1.json
   F-replay-posts-aborted-by-navigation / layer-periodic-events-POST-in-flight run 0:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 1, 'replay events POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 1, 'sessionEvents[net::ERR_ABORTED]': 1}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1, ('sessionReplays/blob', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run0.json
   F-replay-posts-aborted-by-navigation / layer-periodic-events-POST-in-flight run 1:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 1, 'replay events POST (after pagehide)': 1}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 1, 'sessionEvents[net::ERR_ABORTED]': 1}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 3, ('sessionEvents', 200): 1, ('sessionReplays/blob', 200): 1}
       evidence: evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run1.json
   Fr-random-reloads / flush-8-random-reloads-latency300 run 0:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 7, 'replay metadata POST (after pagehide)': 8, 'sdk traces flush (after pagehide)': 2}
       Playwright requestfailed: {'sessionEvents[net::ERR_ABORTED]': 7, 'traces[net::ERR_ABORTED]': 1, 'sessionReplays/meta[net::ERR_ABORTED]': 8}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 17, ('traces', 200): 8, ('sessionEvents', 200): 8, ('sessionReplays/blob', 200): 1}
       evidence: evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run0.json
   Fr-random-reloads / flush-8-random-reloads-latency300 run 1:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 8, 'replay events POST (after pagehide)': 5, 'replay blob PUT (after pagehide)': 1, 'sdk traces flush (after pagehide)': 2}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 8, 'sessionEvents[net::ERR_ABORTED]': 5, 'sessionReplays/blob[net::ERR_ABORTED]': 1, 'traces[net::ERR_ABORTED]': 1}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 17, ('traces', 200): 8, ('sessionEvents', 200): 8, ('sessionReplays/blob', 200): 4}
       evidence: evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run1.json
   Fr-random-reloads / layer-8-random-reloads-latency300 run 0:
       journal (SDK console lines, per document): {'replay events POST (after pagehide)': 7, 'replay metadata POST (after pagehide)': 8}
       Playwright requestfailed: {'sessionEvents[net::ERR_ABORTED]': 7, 'sessionReplays/meta[net::ERR_ABORTED]': 8}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 17, ('sessionEvents', 200): 8, ('traces', 200): 1, ('sessionReplays/blob', 200): 1}
       evidence: evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run0.json
   Fr-random-reloads / layer-8-random-reloads-latency300 run 1:
       journal (SDK console lines, per document): {'replay metadata POST (after pagehide)': 8, 'replay events POST (after pagehide)': 6}
       Playwright requestfailed: {'sessionReplays/meta[net::ERR_ABORTED]': 8, 'sessionEvents[net::ERR_ABORTED]': 6}
       server POSTs answered (path,status): {('sessionReplays/meta', 200): 17, ('sessionEvents', 200): 8, ('traces', 200): 2, ('sessionReplays/blob', 200): 2}
       evidence: evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run1.json
 Notes on these results:
  * In every F run the held, in-flight non-keepalive POST (the periodic
    sessionEvents POST, or the new page's "active" meta row) shows up in Playwright
    as requestfailed net::ERR_ABORTED, and the page prints the matching
    "[maple] session replay events/metadata POST failed (dropping; ...)" line,
    with both presets. Every run also has a second line from the keepalive
    "ended" row, which was delivered (BUG-7).
  * The server answered 200 for every one of these requests, because on localhost
    the body had already arrived before the abort. On a real network an abort
    during upload loses the batch, and the SDK never re-sends it (code above).
  * (Harness note: the first run of the Maple.layer "periodic-events" variant did
    not meet its precondition. Maple.layer starts the replay engine later, so the
    console lines were logged before the console capture existed. The scenario
    now waits for the first "active" row; the runs shown were re-run with that
    fix: evidence/runner-console-F-rerun.txt.)
  * In the Fr runs with the MapleFlush variant, each run also shows one
    "[MapleClientSDK] traces flush failed" + "skipped (cooldown 60000ms
    remaining)" pair after pagehide (BUG-7).
 Caveat: on localhost the mock server had usually already read the request body
 before the abort, so the data arrived anyway. On a real network, an abort during
 upload loses it. What we verified is the abort, the message and the code path
 that drops; we did not measure the loss rate on a real network.

Suggested fixes: restore the batch on failure (bounded) and retry; use keepalive
(within the shared budget) for posts that start after pagehide or while hidden;
fix the message wording; count drops.

------------------------------------------------------------------------------
4.9 BUG-9 / feature gap (INFO): Maple.layer has no unload flush (documented)
------------------------------------------------------------------------------
Documented in the README ("Manual flush", sdk-excerpts/11-readme.txt, L237).
Included for completeness and because a fix is cheap.
Mechanism [VERIFIED-CODE]: Maple.layer = Effect Otlp.layerJson with 5 s export
intervals (client-BYIixzTb.mjs L2286-2330). Effect's exporter (effect
OtlpExporter.js, sdk-excerpts/10-effect-otlp-exporter.txt) also DROPS the batch
on failure and disables export for 60 s (`buffer = []; disabledUntil = now + 60_000`).
Its fetches are not keepalive, so an export in flight at navigation is aborted
and its batch lost.

Empirical [VERIFIED-EMPIRICAL] (scenario H; span "H-early" created 7.5 s before
the reload, "H-tail" 1 s before; evidence/H-maple-layer-unload/*):
      layer          [tail LOST: 3/3]
       run 0: tail LOST | layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
       run 1: tail LOST | layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
       run 2: tail LOST | layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
   layer-flusher  [tail delivered: 3/3]
       run 0: tail delivered | layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
       run 1: tail delivered | layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
       run 2: tail delivered | layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
   flush          [tail delivered: 3/3]
       run 0: tail delivered | flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
       run 1: tail delivered | flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
       run 2: tail delivered | flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true

The "layer-flusher" variant is a few lines of user code (repro/client/app.ts):
   Maple.layer(cfg).pipe(Layer.provide(Layer.succeed(FetchHttpClient.RequestInit, { keepalive: true })))
   + on pagehide / visibilitychange(hidden): runtime.runPromise(Effect.flatMap(OtlpExporter.Flusher, f => f.flush))
Effect 4.0.0-rc.112's Otlp layers already provide OtlpExporter.Flusher (OtlpTracer.js
L94 `Layer.provideMerge(Exporter.layerFlusher)`), so the README statement
"there's no way to force an export" no longer holds with current Effect. Maple.layer
could ship this built in. Caveat: with keepalive on every export it inherits BUG-1/2
unless it chunks and uses keepalive only at unload.


==============================================================================
5. NON-BUGS / ARTIFACTS / DOCUMENTED LIMITATIONS (do not chase these)
==============================================================================

5.1 Playwright "requestfailed net::ERR_ABORTED" for keepalive requests
    [VERIFIED-EMPIRICAL] (scenario G + K + every unload scenario)
      steady-state-no-navigation  [no requestfailed events: 2/2]
       run 0: no requestfailed events | reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
       run 1: no requestfailed events | reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
   with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
       run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) | reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
       run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) | reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
 * Steady state (no navigation): Playwright reports every SDK keepalive request
   as finished, and the server agrees.
 * Any request still in flight when its document is torn down, which includes
   EVERYTHING the pagehide flush sends, is reported as requestfailed
   net::ERR_ABORTED although the server received it and answered 200 (the server
   timestamp precedes the failure event by a few ms). In the browser the same
   requests reject with TypeError (BUG-7).
 * This affects Playwright's view only; delivery is fine. Use server-side
   receipts as the source of truth. The ERR_ABORTED that a real-app dev setup
   reports on "essentially every" blob/traces request is consistent with this
   whenever those requests overlap a navigation (with a dev server, e.g. HMR full
   reloads, navigations are frequent).
   We could NOT reproduce "every request fails even without navigation" in
   either headless mode. Cross-check in full Chromium "new headless":
   full Chromium 'new headless' (evidence/headless-new/):
   steady-state-no-navigation  [no requestfailed events: 2/2]
       run 0: no requestfailed events | reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
       run 1: no requestfailed events | reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
   with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
       run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) | reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
       run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) | reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
   [HYPOTHESIS] Possible reasons in a real-app dev setup: frequent full-page
   reloads or navigations (HMR, auth redirects) so most requests overlap an
   unload; a page whose requests are mostly sent from the pagehide flush; or
   differences in headed Chromium or in Playwright's request tracking when
   service workers or other frames are involved. None of these was tested.

5.2 Pagehide flush works in the simple case [VERIFIED-EMPIRICAL]
 Span created ~150 ms before reload with nothing pending/in flight: delivered
 (C-control 3/3, K 6/6 spans and logs, H "flush" 3/3, I-cross-origin).
 The hypothesis "the pagehide flush never fires / Effect schedules the fetch too
 late" is [DISPROVEN] for Chromium 143.
 [HYPOTHESIS, not observed] The stack traces show the fetch is issued from
 Effect's MixedScheduler ("at MixedScheduler.<anonymous>"). If the Effect
 scheduler ever deferred that work to a macrotask (setTimeout) during pagehide,
 the request would never be issued. Issuing the unload POSTs synchronously in
 the event handler would remove that dependency.

5.3 Cross-origin endpoints [VERIFIED-EMPIRICAL] (scenario I-cross-origin)
 Keepalive + CORS preflight works in Chromium 143 (preflight per request with our
 max-age 0). The preflight asks for "authorization,content-type,x-maple-sdk".
 The SDK also sets a "user-agent" header (version-B225m7RW.mjs L23-27), which
 Chromium silently ignores: the server sees the browser UA. [HYPOTHESIS, untested]
 Browsers that let pages set User-Agent on fetch (Firefox has allowed it for
 years) would add "user-agent" to Access-Control-Request-Headers, so an ingest
 CORS policy that does not list it would block every browser OTLP request in
 those browsers. Consider not sending "user-agent" from the browser entry, since
 x-maple-sdk already carries the same value.
 Runs:
   same-origin  [tail delivered: 2/2]
       run 0: tail delivered | same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
       run 1: tail delivered | same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
   cross-origin  [tail delivered: 2/2]
       run 0: tail delivered | cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
       run 1: tail delivered | cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2

5.4 Session metadata "ended" rows and trace-id lists are capped (200 ids,
 client L1629). We saw "ended" rows of 1.1-6.4 kB. Not a bug; listed because it
 feeds BUG-5.


==============================================================================
6. MISCELLANEOUS SMALLER FINDINGS
==============================================================================

M-1 [VERIFIED-CODE] restore() keeps the OLDEST items when over MAX_BUFFER
    (slice(0, 1e4)) and exportFn drops NEW spans/logs silently at 1e4; no warning,
    no metric (version L301, L313, L443, L485). The type docs mention the 10k cap
    only for autoFlushInterval:false.
M-2 [VERIFIED-EMPIRICAL + CODE] keepaliveFor(requested, body.length) compares a
    JS string length (UTF-16 code units) with a byte budget. A 40,000-character
    body of 2-byte characters is 80,000 bytes and is rejected (B6). Non-ASCII
    console messages, URLs or user names in session events can trip this.
    (0.10.0 fixes this for session posts with TextEncoder.)
M-3 [VERIFIED-EMPIRICAL] The keepalive transport never reads or cancels the
    response body. Chromium keeps counting the request's bytes against the budget
    until the body finishes loading (B3b: back-to-back keepalive POSTs fail unless
    the body is read). Small effect with the "{}" bodies our mock returns, but
    larger or slower response bodies hold budget longer. (0.10.0's postToIngest
    cancels the body; the OTLP transport does not.)
M-4 [VERIFIED-EMPIRICAL] With replay sampled, console.* is wrapped and every
    console call becomes a session event. Effect's default logger (kept by
    Logger.layer([...], {mergeWithExisting:true}), version L305) prints every
    Effect log to the console, so every Effect log is exported twice: as an OTLP
    log and as a "console.log" session event. Evidence: the sessionEvents
    consoleMessages in evidence/D-hang-head-of-line-blocking/*.json contain the
    "INFO (#n): Dhlog-..." lines. This inflates the events payload that competes
    for the unload budget (BUG-5). The SDK's own warnings/errors are captured as
    session events too.
M-5 [VERIFIED-EMPIRICAL] With replay DISABLED, the metadata-only session still
    prints "[maple] session replay events POST failed ..." (K replay-off). The
    wording is confusing for users who turned replay off.
M-6 [VERIFIED-EMPIRICAL] Multi-argument Effect log messages:
    Effect.logError("... Failed to connect", new Error("ClientError: boom"),
    "Unknown Error", {code:42,...}) is POSTed by the SDK with
    logRecord.body = {"arrayValue":{"values":[{"stringValue":"L Failed to connect"},
    {"stringValue":"Error: ClientError: boom"},{"stringValue":"Unknown Error"},
    {"stringValue":"{\"code\":42,\"nested\":{\"a\":1}}"}]}}
    (scenario L, evidence/L-multi-arg-log-body/logError-4-args-run0.json). That
    is valid, structured OTLP (version L338-345 via Effect's
    OtlpResource.unknownToAttributeValue). Our production observation that such
    logs show up in Maple as a JSON-encoded array string
    (`["Failed to connect","ClientError: ...","Unknown Error","{...}"]`) must
    therefore come from how ingest/storage/UI renders an arrayValue body, not
    from the SDK. Ergonomics suggestion: join multiple string parts with " " (as
    Effect's console logger does) or render arrayValue bodies as joined text in
    the UI, with structured args kept as attributes.
M-7 [VERIFIED-CODE] MapleFlush.make() has side effects at construction:
    startClientSession() (L2027-2036) mints the session/visitor ids, writes
    cookies/localStorage, starts the event sink (5 s timer, patched history API,
    click/error listeners) and dynamically imports the replay engine, all before
    (and whether or not) telemetry.layer is ever provided. Calling make() twice
    (e.g. StrictMode/HMR) is guarded by readSessionSink()/lease counting. Not a
    bug, but surprising for a "make" function; worth documenting.
M-8 [VERIFIED-CODE] dispose() awaits flush() before clientSession.stop(), so it
    inherits BUG-3/BUG-4: during cooldown or behind a hung POST, dispose skips or
    stalls.
M-9 [VERIFIED-CODE, investigated, not a bug] Metric buffer: drains a cumulative
    snapshot only when something changed; restore() re-marks dirty. Fine.
    guardFlush: only logs failures around runFlush; fine.
M-10 [HYPOTHESIS] Mobile Safari/Chrome often kill backgrounded tabs without a
    pagehide, after visibilitychange(hidden). There the visibilitychange flush is
    the last chance, and BUG-3/4/5/6 apply equally. Not testable here (no mobile
    browser).


==============================================================================
7. MAPPING TO OUR PRODUCTION OBSERVATIONS
==============================================================================

Obs. 1a: after switching to MapleFlush.make (autoFlushInterval 1000), one of ~6
 Playwright runs logged "[MapleClientSDK] traces flush failed; cooldown 60s:
 TypeError: Failed to fetch" then "[MapleClientSDK] traces flush skipped (cooldown
 60000ms remaining)" around a reload. A later run printed both traces AND logs
 variants around a reload ~150 ms after an action (3/3 runs that day), and the
 span created ~150 ms before the reload still arrived.
 -> Reproduced as scenario K: 6/6 runs show the traces and logs variants;
    span + log delivered in 6/6. Attributed to the OLD document's pagehide flush
    (failed) and visibilitychange flush (skipped, "60000ms remaining"); the data
    was delivered (BUG-7). It is intermittent in production because it needs
    something buffered at pagehide: with autoFlushInterval 1000 the buffer is
    usually empty unless the action happened within ~1 s of the reload.
    Concurrent budget (BUG-2/5) was the leading hypothesis for this pattern.
    [DISPROVEN] for these small payloads: the server got every request. It
    remains possible with large payloads, which print the same line.
Obs. 1b: in one run, a span that ended ~5 s before a reload never arrived (its
 sibling that ended earlier did); not reproducible on rerun.
 -> Most consistent with BUG-4: one slow ingest-proxy round-trip (several
    seconds) blocks every later flush, including the pagehide flush. Reproduced
    with a 7 s round-trip and a span ending 5 s before the reload ([REPRODUCED (tail span lost): 3/3]
       run 0: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
       run 1: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
       run 2: REPRODUCED (tail span lost) | first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms).
    Other ways to lose it (BUG-1/2/3/5) would have printed "flush failed"
    BEFORE the reload. In the BUG-4 case the console at the reload shows one
    "traces flush failed; cooldown 60s: TypeError: Failed to fetch" followed by
    one or more "traces flush skipped (cooldown 5999xms remaining)" lines; the
    number of "skipped" lines is the number of flushes that were queued. If the
    production console of that run showed more than one "skipped" line at the
    reload, that points to BUG-4. [HYPOTHESIS for the production case.]
Obs. 2: with BOTH presets, "[maple] session replay events/metadata POST failed
 (dropping; will retry on next chunk): TypeError: Failed to fetch" around
 navigations; Playwright showed ERR_ABORTED for sessionEvents and
 sessionReplays/meta at those moments.
 -> Mostly BUG-7 (delivered, document gone); reproduced at every reload with
    pending session data (K, C-control, E, F). Real loss is possible through
    BUG-8 (aborted non-keepalive periodic posts) and BUG-5 (budget). The final
    replay chunk is lost at every unload anyway (BUG-6), with no message at all.
Obs. 3: Playwright reports essentially every keepalive request
 (traces, sessionReplays/blob) as requestfailed ERR_ABORTED although the data
 arrived.
 -> Section 5.1: reproduced for every request in flight when its document goes
    away. We did NOT see it for steady-state requests in headless-shell or new
    headless Chromium 143 without navigation.
Obs. 4: steady state without navigation: everything arrives; send -> 150 ms ->
 reload delivered the final span.
 -> Consistent (G steady, C-control, K, H flush).


==============================================================================
8. CHECK AGAINST THE LATEST SDK (0.10.0, published 2026-10-03)
==============================================================================

Code reading [VERIFIED-CODE] (sdk-excerpts/12-v0.10.0-what-changed-and-what-did-not.txt):
 * NEW: a shared page-wide keepalive budget for session/replay posts
   (KEEPALIVE_BUDGET_BYTES = 49152 on globalThis, reserve/release, TextEncoder
   byte length, response body cancelled before release). This addresses part of
   BUG-5 (meta/events/blob among themselves) and M-2/M-3 for those posts.
 * UNCHANGED: MapleFlush keepaliveTransport (always keepalive:true, no
   reservation, no size check), flushSignal (whole batch, restore, 60 s cooldown,
   cooldown also at unload), runFlush (Promise.all over three signals), and
   makeSerializedFlush for the browser (the new coalesceSameArguments option is
   documented as a Workers option). So BUG-1/2/3/4 remain. The new comment says
   the session budget "stops short of 64 KiB because the OTLP trace exporter
   spends from the same page-wide allowance under its own accounting", but the
   OTLP transport has no accounting; an OTLP body above ~16 KiB at unload still
   starves the session posts, or the reverse (BUG-5 remains).
 * UNCHANGED: the replay uploadChunk() claims the chunk seq, then awaits gzip,
   then POSTs (BUG-6 remains).
 * UNCHANGED: warnDropped wording (BUG-8).
Empirical (same scenarios, SDK=0.10.0 build, effect 4.0.0; evidence/sdk-0.10.0/):
      A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
   A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
   A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
   A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
   A0-threshold / n100  [DELIVERED: 2/2]
   A0-threshold / n150  [DELIVERED: 2/2]
   A0-threshold / n160  [DELIVERED: 2/2]
   A0-threshold / n163  [DELIVERED: 2/2]
   A0-threshold / n164  [REJECTED: 2/2]
   A0-threshold / n165  [REJECTED: 2/2]
   A0-threshold / n170  [REJECTED: 2/2]
   A0-threshold / n200  [REJECTED: 2/2]
   B-probe / probe  [MEASURED: 2/2]
   B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
   B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
   C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 2/2]
   C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 2/2]
   C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 2/2]
   C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 2/2]
   D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
   D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 2/2]
   D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 2/2]
   D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 2/2]
   D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 2/2]
   E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
   E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
   E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
   K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 2/2]
   (per-run details: evidence/sdk-0.10.0/summary-*.txt and the per-run JSON files)


==============================================================================
9. SUGGESTED FIXES (consolidated, roughly by priority)
==============================================================================
 1. OTLP transport: chunk by serialized UTF-8 size (<= 48 KiB per request);
    steady-state POSTs without keepalive (with a timeout); keepalive only on
    the unload path. (BUG-1, BUG-2, BUG-4)
 2. One page-wide keepalive budget object shared by OTLP + session/replay posts
    and by multiple SDK copies; reserve on send, release when the response body
    ends; fall back to non-keepalive when a reservation fails; prioritise at
    unload. (BUG-2, BUG-5, M-2, M-3)
 3. Unload path: synchronous drain + send inside the pagehide/visibilitychange
    task, never blocked by the serialized chain, never skipped by the cooldown,
    and no cooldown/restore/console.error for failures after pagehide. (BUG-3,
    BUG-4, BUG-7)
 4. On a keepalive TypeError at send time: retry without keepalive or with smaller
    chunks; do not restore an unsendable batch forever; keep the newest data
    when trimming. (BUG-1, M-1)
 5. Replay: have the final chunk ready synchronously at pagehide (incremental
    gzip or uncompressed fallback) or persist it for the next load; claim the seq
    only when sending. (BUG-6)
 6. Session posts: restore + retry instead of drop, fix the "will retry" wording,
    throttle less aggressively or count drops. (BUG-8, M-5)
 7. Observability of the SDK itself: counters for dropped spans/logs/events by
    reason (buffer cap, oversized, keepalive rejected, cooldown at unload),
    exposed via an API or as a metric. Today the only signal is a console line
    that is also printed for successful requests.
 8. Optional sendBeacon fallback for the unload path if ingest accepted auth via
    query parameter or body. Beacons have the same 64 KiB budget in Chromium but
    are queued even when the document dies mid-flush. Not needed for
    same-origin-proxy setups like ours.
 9. Maple.layer: register a pagehide/visibilitychange hook that calls Effect's
    OtlpExporter.Flusher with keepalive (with chunking), as in 4.9.
10. Docs: state the 64 KiB keepalive budget, the per-flush span limit (~160 plain
    spans), and that a single failure skips the unload flush for 60 s (until
    fixed).


==============================================================================
APPENDIX A. FULL SCENARIO TABLE (Chromium 143 headless-shell, SDK 0.8.2)
==============================================================================
(Generated from evidence/summary-*.txt; each line = scenario / variant [verdict counts].)

--- SDK 0.8.2, chromium-headless-shell (evidence/) ---
B-probe / probe  [MEASURED: 2/2]
   run 0: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/B-keepalive-probe/probe-run0.json
   run 1: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/B-keepalive-probe/probe-run1.json
B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 3/3]
   run 0: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/manual-flush-run0.json
   run 1: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/manual-flush-run1.json
   run 2: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/manual-flush-run2.json
B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 3/3]
   run 0: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/reload-run0.json
   run 1: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/reload-run1.json
   run 2: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/B-sdk-concurrent-signals/reload-run2.json
A0-threshold / n100  [DELIVERED: 1/1]
   run 0: 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: evidence/A0-threshold/n100-run0.json
A0-threshold / n150  [DELIVERED: 1/1]
   run 0: 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: evidence/A0-threshold/n150-run0.json
A0-threshold / n160  [DELIVERED: 1/1]
   run 0: 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: evidence/A0-threshold/n160-run0.json
A0-threshold / n163  [DELIVERED: 1/1]
   run 0: 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: evidence/A0-threshold/n163-run0.json
A0-threshold / n164  [REJECTED: 1/1]
   run 0: 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/A0-threshold/n164-run0.json
A0-threshold / n165  [REJECTED: 1/1]
   run 0: 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/A0-threshold/n165-run0.json
A0-threshold / n170  [REJECTED: 1/1]
   run 0: 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/A0-threshold/n170-run0.json
A0-threshold / n200  [REJECTED: 1/1]
   run 0: 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/A0-threshold/n200-run0.json
C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 3/3]
   run 0: inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2937,t=2942
      evidence: evidence/C-unload-during-cooldown/inject503-then-reload-run0.json
   run 1: inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2947,t=2956
      evidence: evidence/C-unload-during-cooldown/inject503-then-reload-run1.json
   run 2: inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2122,t=2952,t=2958
      evidence: evidence/C-unload-during-cooldown/inject503-then-reload-run2.json
A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
   run 1: created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
   run 0: created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json
A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
   run 1: created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
   run 0: created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json
C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 3/3]
   run 0: inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2937,t=2941
      evidence: evidence/C-unload-during-cooldown/inject503-then-navigate-run0.json
   run 1: inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2108,t=2937,t=2943
      evidence: evidence/C-unload-during-cooldown/inject503-then-navigate-run1.json
   run 2: inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2147,t=2960,t=2965
      evidence: evidence/C-unload-during-cooldown/inject503-then-navigate-run2.json
C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 3/3]
   run 0: inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2938,t=2942
      evidence: evidence/C-unload-during-cooldown/inject503-then-close-run0.json
   run 1: inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2930,t=2935
      evidence: evidence/C-unload-during-cooldown/inject503-then-close-run1.json
   run 2: inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2939,t=2945
      evidence: evidence/C-unload-during-cooldown/inject503-then-close-run2.json
A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
   run 1: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2565,63567,124567ms
      evidence: evidence/A-oversized-batch/burst200x500B-interval1000-run1.json
   run 0: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2548,63547,124547ms
      evidence: evidence/A-oversized-batch/burst200x500B-interval1000-run0.json
A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
   run 0: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5491,70488,135493ms
      evidence: evidence/A-oversized-batch/burst200x500B-interval5000(default)-run0.json
   run 1: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5483,70483,135481ms
      evidence: evidence/A-oversized-batch/burst200x500B-interval5000(default)-run1.json
C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 3/3]
   run 0: inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2972
      evidence: evidence/C-unload-during-cooldown/control-no-failure-reload-run0.json
   run 1: inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2949
      evidence: evidence/C-unload-during-cooldown/control-no-failure-reload-run1.json
   run 2: inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2970
      evidence: evidence/C-unload-during-cooldown/control-no-failure-reload-run2.json
D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 3/3]
   run 0: first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5125ms); D-tail delivered=false; reload at t=1401ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
   run 1: first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5107ms); D-tail delivered=false; reload at t=1382ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json
   run 2: first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5108ms); D-tail delivered=false; reload at t=1375ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run2.json
D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 3/3]
   run 0: first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9124ms); D-tail delivered=false; reload at t=5383ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
   run 1: first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9115ms); D-tail delivered=false; reload at t=5367ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json
   run 2: first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9114ms); D-tail delivered=false; reload at t=5373ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run2.json
D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 3/3]
   run 0: first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
   run 1: first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json
   run 2: first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms
      evidence: evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run2.json
D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 3/3]
   run 0: first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1405ms); reload at t=1393ms
      evidence: evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
   run 1: first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1412ms); reload at t=1406ms
      evidence: evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json
   run 2: first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1123ms); D-tail delivered=true (received t=1401ms); reload at t=1394ms
      evidence: evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run2.json
D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
   run 0: first traces POST held 15000 ms (from t=1153ms): span export delay max 14990 ms / min 909 ms; log export delay max 14991 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17143:1sp/1157B 18143:1sp/1157B 19144:1sp/1157B]
      evidence: evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
   run 1: first traces POST held 15000 ms (from t=1137ms): span export delay max 14977 ms / min 905 ms; log export delay max 14977 ms (logs endpoint itself is fast); traces POSTs at server t=[1116:1sp/1160B 16120:15sp/6677B 17115:1sp/1157B 18115:1sp/1157B 19115:1sp/1157B]
      evidence: evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json
Dr-random-reload-latency / latency150-interval1000  [no loss: 9/10; LOSS (1 spans): 1/10]
   run 0: latency 150 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run0.json
   run 1: latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run1.json
   run 2: latency 150 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run2.json
   run 3: latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run3.json
   run 4: latency 150 ms, interval 1000: created 38, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run4.json
   run 5: latency 150 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run5.json
   run 6: latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run6.json
   run 7: latency 150 ms, interval 1000: created 40, lost 1 (created 100 ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run7.json
   run 8: latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run8.json
   run 9: latency 150 ms, interval 1000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency150-interval1000-run9.json
Dr-random-reload-latency / latency600-interval1000  [no loss: 7/10; LOSS (5 spans): 1/10; LOSS (2 spans): 1/10; LOSS (4 spans): 1/10]
   run 0: latency 600 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run0.json
   run 1: latency 600 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run1.json
   run 2: latency 600 ms, interval 1000: created 54, lost 5 (created 513,411,305,202,100 ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run2.json
   run 3: latency 600 ms, interval 1000: created 22, lost 2 (created 202,100 ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run3.json
   run 4: latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run4.json
   run 5: latency 600 ms, interval 1000: created 57, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run5.json
   run 6: latency 600 ms, interval 1000: created 39, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run6.json
   run 7: latency 600 ms, interval 1000: created 34, lost 4 (created 406,304,202,100 ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run7.json
   run 8: latency 600 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run8.json
   run 9: latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval1000-run9.json
Dr-random-reload-latency / latency600-interval5000  [no loss: 10/10]
   run 0: latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run0.json
   run 1: latency 600 ms, interval 5000: created 41, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run1.json
   run 2: latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run2.json
   run 3: latency 600 ms, interval 5000: created 22, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run3.json
   run 4: latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run4.json
   run 5: latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run5.json
   run 6: latency 600 ms, interval 5000: created 35, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run6.json
   run 7: latency 600 ms, interval 5000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run7.json
   run 8: latency 600 ms, interval 5000: created 21, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run8.json
   run 9: latency 600 ms, interval 5000: created 25, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/latency600-interval5000-run9.json
Dr-random-reload-latency / control-latency0-interval1000  [no loss: 10/10]
   run 0: latency 0 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run0.json
   run 1: latency 0 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run1.json
   run 2: latency 0 ms, interval 1000: created 43, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run2.json
   run 3: latency 0 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run3.json
   run 4: latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run4.json
   run 5: latency 0 ms, interval 1000: created 31, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run5.json
   run 6: latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run6.json
   run 7: latency 0 ms, interval 1000: created 33, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run7.json
   run 8: latency 0 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run8.json
   run 9: latency 0 ms, interval 1000: created 23, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: evidence/Dr-random-reload-latency/control-latency0-interval1000-run9.json
E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
   run 0: spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
   run 1: spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json
   run 2: spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run2.json
E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 3/3]
   run 0: spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
   run 1: spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json
   run 2: spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run2.json
E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 3/3]
   run 0: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
   run 1: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json
   run 2: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run2.json
E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json
   run 2: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run2.json
E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json
   run 2: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run2.json
F-replay-posts-aborted-by-navigation / flush-active-meta-POST-in-flight  [REPRODUCED: 2/2]
   run 0: flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run0.json
   run 1: flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run1.json
Fr-random-reloads / flush-8-random-reloads-latency300  [REPRODUCED: 2/2]
   run 0: flush, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"traces[net::ERR_ABORTED]":1,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run0.json
   run 1: flush, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST; [maple] session replay blob PUT], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":5,"sessionReplays/blob[net::ERR_ABORTED]":1,"traces[net::ERR_ABORTED]":1}
      evidence: evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run1.json
F-replay-posts-aborted-by-navigation / layer-active-meta-POST-in-flight  [REPRODUCED: 2/2]
   run 0: layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run0.json
   run 1: layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run1.json
Fr-random-reloads / layer-8-random-reloads-latency300  [REPRODUCED: 2/2]
   run 0: layer, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run0.json
   run 1: layer, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":6}
      evidence: evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run1.json
G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
   run 0: reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
   run 1: reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json
G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
   run 0: reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/G-playwright-keepalive-artifact/with-final-reload-run0.json
   run 1: reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/G-playwright-keepalive-artifact/with-final-reload-run1.json
H-maple-layer-unload / layer  [tail LOST: 3/3]
   run 0: layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: evidence/H-maple-layer-unload/layer-run0.json
   run 1: layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: evidence/H-maple-layer-unload/layer-run1.json
   run 2: layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: evidence/H-maple-layer-unload/layer-run2.json
H-maple-layer-unload / layer-flusher  [tail delivered: 3/3]
   run 0: layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/layer-flusher-run0.json
   run 1: layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/layer-flusher-run1.json
   run 2: layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/layer-flusher-run2.json
H-maple-layer-unload / flush  [tail delivered: 3/3]
   run 0: flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/flush-run0.json
   run 1: flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/flush-run1.json
   run 2: flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: evidence/H-maple-layer-unload/flush-run2.json
I-cross-origin / same-origin  [tail delivered: 2/2]
   run 0: same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: evidence/I-cross-origin/same-origin-run0.json
   run 1: same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: evidence/I-cross-origin/same-origin-run1.json
I-cross-origin / cross-origin  [tail delivered: 2/2]
   run 0: cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: evidence/I-cross-origin/cross-origin-run0.json
   run 1: cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: evidence/I-cross-origin/cross-origin-run1.json
K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 3/3]
   run 0: reload at t=1795; requests issued around the reload (Playwright): traces(1165B)@1807 logs(1014B)@1807 sessionReplays/meta(1164B)@1807 sessionEvents(1021B)@1807 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1801 logs@1802 sessionReplays/meta@1802 sessionEvents@1804 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flu
      evidence: evidence/K-reload-150ms-after-action/replay-on(default)-run0.json
   run 1: reload at t=1810; requests issued around the reload (Playwright): traces(1165B)@1824 logs(1014B)@1824 sessionReplays/meta(1164B)@1824 sessionEvents(1021B)@1824 sessionReplays/meta(1060B)@1891; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1817 sessionReplays/meta@1818 logs@1818 sessionEvents@1820 sessionReplays/meta@1890; console: [MapleClientSDK] logs flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flu
      evidence: evidence/K-reload-150ms-after-action/replay-on(default)-run1.json
   run 2: reload at t=1803; requests issued around the reload (Playwright): traces(1165B)@1815 logs(1014B)@1816 sessionReplays/meta(1164B)@1816 sessionEvents(1021B)@1816 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1809 logs@1810 sessionReplays/meta@1813 sessionEvents@1814 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flu
      evidence: evidence/K-reload-150ms-after-action/replay-on(default)-run2.json
K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 3/3]
   run 0: reload at t=1770; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1781 sessionEvents(499B)@1781 traces(1165B)@1782 logs(1014B)@1782 sessionReplays/meta(1070B)@1814; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1777 traces@1778 sessionEvents@1779 logs@1781 sessionReplays/meta@1811; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1787 | [MapleClientSDK] traces flus
      evidence: evidence/K-reload-150ms-after-action/replay-off-run0.json
   run 1: reload at t=1771; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1783 sessionEvents(499B)@1783 traces(1165B)@1784 logs(1014B)@1784 sessionReplays/meta(1070B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1778 sessionEvents@1779 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1785 | [MapleClientSDK] traces flus
      evidence: evidence/K-reload-150ms-after-action/replay-off-run1.json
   run 2: reload at t=1769; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1780 sessionEvents(499B)@1780 traces(1165B)@1781 logs(1014B)@1781 sessionReplays/meta(1070B)@1813; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1776 traces@1777 sessionEvents@1778 logs@1780 sessionReplays/meta@1810; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1782 | [MapleClientSDK] traces flus
      evidence: evidence/K-reload-150ms-after-action/replay-off-run2.json
L-multi-arg-log-body / logError-4-args  [body is an OTLP arrayValue (structured), not a JSON string: 1/1]
   run 0: OTLP logRecord.body as POSTed by the SDK: {"arrayValue":{"values":[{"stringValue":"L Failed to connect"},{"stringValue":"Error: ClientError: boom"},{"stringValue":"Unknown Error"},{"stringValue":"{\"code\":42,\"nested\":{\"a\":1}}"}]}}
      evidence: evidence/L-multi-arg-log-body/logError-4-args-run0.json
W-workaround-keepalive-downgrade-shim / without-shim  [LOSS (burst 0/200, ticks 0/5): 2/2]
   run 0: shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: evidence/W-workaround-keepalive-downgrade-shim/without-shim-run0.json
   run 1: shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: evidence/W-workaround-keepalive-downgrade-shim/without-shim-run1.json
W-workaround-keepalive-downgrade-shim / with-keepalive-downgrade-shim  [all delivered: 2/2]
   run 0: shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run0.json
   run 1: shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run1.json
F-replay-posts-aborted-by-navigation / layer-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
   run 0: layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run0.json
   run 1: layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run1.json
F-replay-posts-aborted-by-navigation / flush-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
   run 0: flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run0.json
   run 1: flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run1.json

--- SDK 0.10.0 + effect 4.0.0, chromium-headless-shell (evidence/sdk-0.10.0/) ---
A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
   run 1: created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10460,75460ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
   run 0: created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10481,75482ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json
A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
   run 1: created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
   run 0: created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json
A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
   run 1: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2447,63447,124447ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run1.json
   run 0: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2513,63514,124513ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run0.json
A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
   run 1: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5451,70449,135456ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run1.json
   run 0: created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5488,70485,135492ms
      evidence: evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run0.json
B-probe / probe  [MEASURED: 2/2]
   run 0: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/sdk-0.10.0/B-keepalive-probe/probe-run0.json
   run 1: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/sdk-0.10.0/B-keepalive-probe/probe-run1.json
B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
   run 0: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run0.json
   run 1: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run1.json
B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
   run 0: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run0.json
   run 1: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run1.json
A0-threshold / n100  [DELIVERED: 2/2]
   run 0: 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n100-run0.json
   run 1: 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n100-run1.json
A0-threshold / n150  [DELIVERED: 2/2]
   run 0: 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n150-run0.json
   run 1: 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n150-run1.json
A0-threshold / n160  [DELIVERED: 2/2]
   run 0: 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n160-run0.json
   run 1: 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n160-run1.json
A0-threshold / n163  [DELIVERED: 2/2]
   run 0: 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n163-run0.json
   run 1: 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: evidence/sdk-0.10.0/A0-threshold/n163-run1.json
A0-threshold / n164  [REJECTED: 2/2]
   run 0: 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n164-run0.json
   run 1: 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n164-run1.json
A0-threshold / n165  [REJECTED: 2/2]
   run 0: 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n165-run0.json
   run 1: 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n165-run1.json
A0-threshold / n170  [REJECTED: 2/2]
   run 0: 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n170-run0.json
   run 1: 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n170-run1.json
A0-threshold / n200  [REJECTED: 2/2]
   run 0: 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n200-run0.json
   run 1: 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/sdk-0.10.0/A0-threshold/n200-run1.json
C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 2/2]
   run 0: inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2115,t=2950,t=2956
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run0.json
   run 1: inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2949,t=2953
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run1.json
C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 2/2]
   run 0: inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2119,t=2954,t=2960
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run0.json
   run 1: inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2956,t=2962
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run1.json
C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 2/2]
   run 0: inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2931,t=2936
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run0.json
   run 1: inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2134,t=2949,t=2959
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run1.json
C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 2/2]
   run 0: inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2960
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run0.json
   run 1: inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2954
      evidence: evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run1.json
D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 2/2]
   run 0: first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5114ms); D-tail delivered=false; reload at t=1383ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
   run 1: first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5124ms); D-tail delivered=false; reload at t=1380ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json
D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 2/2]
   run 0: first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9123ms); D-tail delivered=false; reload at t=5383ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
   run 1: first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9116ms); D-tail delivered=false; reload at t=5370ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json
D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 2/2]
   run 0: first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8113ms); D-tail delivered=false; reload at t=6231ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
   run 1: first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6241ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json
D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 2/2]
   run 0: first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1145ms); D-tail delivered=true (received t=1408ms); reload at t=1401ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
   run 1: first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1115ms); D-tail delivered=true (received t=1396ms); reload at t=1390ms
      evidence: evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json
D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
   run 0: first traces POST held 15000 ms (from t=1161ms): span export delay max 14983 ms / min 886 ms; log export delay max 14983 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17144:1sp/1157B 18144:1sp/1157B 19144:1sp/1157B]
      evidence: evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
   run 1: first traces POST held 15000 ms (from t=1131ms): span export delay max 14992 ms / min 922 ms; log export delay max 14992 ms (logs endpoint itself is fast); traces POSTs at server t=[1123:1sp/1160B 16128:15sp/6677B 17123:1sp/1157B 18123:1sp/1157B 19122:1sp/1157B]
      evidence: evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json
E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
   run 1: spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json
E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
   run 0: spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
   run 1: spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json
E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
   run 0: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
   run 1: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json
E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json
E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json
K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 2/2]
   run 0: reload at t=1777; requests issued around the reload (Playwright): traces(1165B)@1789 logs(1014B)@1790 sessionReplays/meta(1194B)@1790 sessionEvents(1043B)@1790 sessionReplays/meta(1090B)@1850; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1783 sessionReplays/meta@1785 logs@1785 sessionEvents@1787 sessionReplays/meta@1849; console: [MapleClientSDK] logs flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush failed; cooldown 
      evidence: evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run0.json
   run 1: reload at t=1807; requests issued around the reload (Playwright): traces(1165B)@1819 logs(1014B)@1819 sessionReplays/meta(1194B)@1819 sessionEvents(1043B)@1819 sessionReplays/meta(1090B)@1882; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1813 logs@1814 sessionReplays/meta@1815 sessionEvents@1816 sessionReplays/meta@1882; console: [MapleClientSDK] logs flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush failed; cooldown 
      evidence: evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run1.json
K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
   run 0: reload at t=1783; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1792 sessionEvents(510B)@1793 traces(1165B)@1793 logs(1014B)@1793 sessionReplays/meta(1100B)@1825; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1789 sessionEvents@1790 traces@1790 logs@1791 sessionReplays/meta@1822; console: [MapleClientSDK] traces flush failed; cooldown 60s@1793 | [MapleClientSDK] logs flush failed; cooldown 6
      evidence: evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run0.json
   run 1: reload at t=1772; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1782 sessionEvents(510B)@1783 traces(1165B)@1783 logs(1014B)@1784 sessionReplays/meta(1100B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1779 sessionEvents@1780 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 6
      evidence: evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run1.json

--- SDK 0.8.2, full Chromium new headless (evidence/headless-new/) ---
B-probe / probe  [MEASURED: 2/2]
   run 0: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/headless-new/B-keepalive-probe/probe-run0.json
   run 1: single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: evidence/headless-new/B-keepalive-probe/probe-run1.json
B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
   run 0: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run0.json
   run 1: trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run1.json
B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
   run 0: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/headless-new/B-sdk-concurrent-signals/reload-run0.json
   run 1: trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: evidence/headless-new/B-sdk-concurrent-signals/reload-run1.json
E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
   run 1: spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json
E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
   run 0: spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
   run 1: spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json
E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
   run 0: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
   run 1: spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json
E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json
E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
   run 0: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
   run 1: spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json
G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
   run 0: reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
   run 1: reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json
G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
   run 0: reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run0.json
   run 1: reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run1.json
K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 1/2; span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 1/2]
   run 0: reload at t=1856; requests issued around the reload (Playwright): traces(1162B)@1871 logs(1011B)@1871 sessionReplays/meta(1160B)@1872 sessionEvents(1019B)@1872 sessionReplays/meta(1056B)@1931; Playwright failed: sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1865 logs@1867 sessionReplays/meta@1867 sessionEvents@1868 sessionReplays/meta@1930; console: [MapleClientSDK] logs flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flu
      evidence: evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run0.json
   run 1: reload at t=1880; requests issued around the reload (Playwright): traces(1162B)@1899 logs(1011B)@1899 sessionReplays/meta(1160B)@1900 sessionEvents(1019B)@1900 sessionReplays/meta(1056B)@1964; Playwright failed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: traces@1890 logs@1891 sessionReplays/meta@1893 sessionEvents@1895 sessionReplays/meta@1963; console: [MapleClientSDK] traces flush failed; cooldown 60s@1901 | [MapleClientSDK] logs flush failed; cooldown 60s@1901 | [MapleClientSDK] traces flu
      evidence: evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run1.json
K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
   run 0: reload at t=1833; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1846 sessionEvents(498B)@1847 traces(1162B)@1847 logs(1011B)@1847 sessionReplays/meta(1066B)@1888; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1842 traces@1844 sessionEvents@1844 logs@1845 sessionReplays/meta@1886; console: [MapleClientSDK] traces flush failed; cooldown 60s@1852 | [MapleClientSDK] logs flush failed; cooldown 60s@1852 | [MapleClientSDK] traces flus
      evidence: evidence/headless-new/K-reload-150ms-after-action/replay-off-run0.json
   run 1: reload at t=1819; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1831 sessionEvents(498B)@1831 traces(1162B)@1831 logs(1011B)@1831 sessionReplays/meta(1066B)@1868; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1827 sessionEvents@1828 traces@1828 logs@1830 sessionReplays/meta@1867; console: [MapleClientSDK] traces flush failed; cooldown 60s@1832 | [MapleClientSDK] logs flush failed; cooldown 60s@1832 | [MapleClientSDK] traces flus
      evidence: evidence/headless-new/K-reload-150ms-after-action/replay-off-run1.json



==============================================================================
APPENDIX B. RAW LOG SNIPPETS
==============================================================================

Extracted verbatim from the evidence JSON files (t values = ms since the start of
that run; 'server recv/resp' = when the mock ingest got the request / sent its
response; console lines truncated to their first line, full text including stack
traces in the JSON).

### BUG-1: burst of 200 spans x 500 B, autoFlushInterval 1000 (no /v1/traces POST ever reaches the server; logs do)
    file: evidence/A-oversized-batch/burst200x500B-interval1000-run0.json
    marks: goto@272, ready@651, burst-created@1674
    console t=  2548 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    console t= 63547 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    console t=124547 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch

### BUG-1: 164 plain spans in one flush (65,596 B) -> rejected
    file: evidence/A0-threshold/n164-run0.json
    marks: goto@39, ready@212, spans-created@230, flush-returned@246
    console t=   246 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    playwright t=   212 request  POST sessionReplays/meta body=1045B
    playwright t=   245 request  POST traces body=65596B
    playwright t=   263 finished POST sessionReplays/meta status=200
    server recv=   210 resp=210 POST sessionReplays/meta 1045B status=200 rows=[('active', 1)]

### BUG-2: traces 38,472 B + logs 32,135 B + metrics 931 B in one flush
    file: evidence/B-sdk-concurrent-signals/manual-flush-run0.json
    marks: goto@30, ready@168, data-created@190, flush-returned@202
    console t=   202 error   [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    playwright t=   167 request  POST sessionReplays/meta body=1057B
    playwright t=   199 request  POST traces body=38472B
    playwright t=   200 request  POST logs body=32135B
    playwright t=   201 request  POST metrics body=931B
    playwright t=   219 finished POST sessionReplays/meta status=200
    playwright t=   253 finished POST traces status=200
    playwright t=   253 finished POST metrics status=200
    server recv=   166 resp=167 POST sessionReplays/meta 1057B status=200 rows=[('active', 1)]
    server recv=   195 resp=195 POST traces 38472B status=200 spans=40 ['Bsdk-span-0', 'Bsdk-span-1', 'Bsdk-span-2', 'Bsdk-span-3']...
    server recv=   197 resp=197 POST metrics 931B status=200 

### BUG-3: one 503, then reload 1.8 s later
    file: evidence/C-unload-during-cooldown/inject503-then-reload-run0.json
    marks: goto@33, ready@158, C-first created@162, C-first POST received by server@1127, C-before-unload created@2630, reload-start@2930, reload-load-event@2951, reload-ready@2989
    console t=  1126 error   [MapleClientSDK] traces flush failed; cooldown 60s: Error: OTLP 503 Service Unavailable
    console t=  2120 warning [MapleClientSDK] traces flush skipped (cooldown 59005ms remaining)
    console t=  2937 warning [MapleClientSDK] traces flush skipped (cooldown 58188ms remaining)
    console t=  2942 warning [MapleClientSDK] traces flush skipped (cooldown 58184ms remaining)
    playwright t=   158 request  POST sessionReplays/meta body=1056B
    playwright t=   209 finished POST sessionReplays/meta status=200
    playwright t=  1123 request  POST traces body=1159B
    playwright t=  1175 finished POST traces status=503
    playwright t=  2941 request  POST sessionReplays/meta body=1195B
    playwright t=  2942 request  POST sessionEvents body=2167B
    playwright t=  2943 failed   POST sessionEvents [net::ERR_ABORTED]
    playwright t=  2943 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  3012 request  POST sessionReplays/meta body=1056B
    playwright t=  3064 finished POST sessionReplays/meta status=200
    server recv=   157 resp=157 POST sessionReplays/meta 1056B status=200 rows=[('active', 1)]
    server recv=  1122 resp=1123 POST traces 1159B status=503 spans=1 ['C-first']
    server recv=  2938 resp=2938 POST sessionReplays/meta 1195B status=200 rows=[('ended', 2)]
    server recv=  2940 resp=2940 POST sessionEvents 2167B status=200 rows=4 seqs=[0, 1, 2, 3]
    server recv=  3012 resp=3012 POST sessionReplays/meta 1056B status=200 rows=[('active', 3)]

### BUG-4: first traces POST held 7 s; D-tail ends 5 s before reload; never sent
    file: evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
    marks: goto@28, ready@110, D-early created@135, D-early POST received by server (response held for latencyMs)@1121, D-tail created@1226, reload-start@6226, reload-load-event@6244, reload-ready@6270
    console t=  6236 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    console t=  6236 warning [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
    console t=  6236 warning [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
    console t=  6236 warning [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
    console t=  6236 warning [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
    console t=  6236 warning [MapleClientSDK] traces flush skipped (cooldown 59998ms remaining)
    console t=  6237 warning [MapleClientSDK] traces flush skipped (cooldown 59998ms remaining)
    console t=  6237 warning [MapleClientSDK] traces flush skipped (cooldown 59998ms remaining)
    journal t=    83 script-start {"docId": "vv82k0", "variant": "flush", "origin": "same", "autoFlushInterval": 1000, "replayEnabled": true}
    journal t=   106 runtime-ready {}
    journal t=   134 span-created {"name": "D-early"}
    journal t=  1224 span-created {"name": "D-tail"}
    journal t=  6231 pagehide(page listener, registered before SDK) {"docId": "vv82k0", "persisted": false}
    journal t=  6232 visibilitychange:hidden {"docId": "vv82k0"}
    journal t=  6233 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: F
    journal t=  6234 console.error {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch"}
    journal t=  6235 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining) "}
    journal t=  6235 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining) "}
    journal t=  6235 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining) "}
    journal t=  6235 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining) "}
    journal t=  6236 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59998ms remaining) "}
    journal t=  6236 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59998ms remaining) "}
    journal t=  6236 console.warn {"docId": "vv82k0", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59998ms remaining) "}
    journal t=  6266 script-start {"docId": "s6rfd3", "variant": "flush", "origin": "same", "autoFlushInterval": 1000, "replayEnabled": true}
    journal t=  6268 runtime-ready {}
    playwright t=   133 request  POST sessionReplays/meta body=1064B
    playwright t=   186 finished POST sessionReplays/meta status=200
    playwright t=  1107 request  POST traces body=1159B
    playwright t=  5088 request  POST sessionEvents body=494B
    playwright t=  5134 request  POST sessionReplays/blob body=623B
    playwright t=  5140 finished POST sessionEvents status=200
    playwright t=  5186 finished POST sessionReplays/blob status=200
    playwright t=  6235 request  POST sessionReplays/meta body=1203B
    playwright t=  6235 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  6235 failed   POST traces [net::ERR_ABORTED]
    playwright t=  6295 request  POST sessionReplays/meta body=1064B
    playwright t=  6346 finished POST sessionReplays/meta status=200
    playwright t= 11269 request  POST sessionEvents body=494B
    playwright t= 11295 request  POST sessionReplays/blob body=623B
    playwright t= 11320 finished POST sessionEvents status=200
    playwright t= 11346 finished POST sessionReplays/blob status=200
    server recv=   133 resp=133 POST sessionReplays/meta 1064B status=200 rows=[('active', 1)]
    server recv=  1106 resp=8107 POST traces 1159B status=200 spans=1 ['D-early']
    server recv=  5088 resp=5088 POST sessionEvents 494B status=200 rows=1 seqs=[0]
    server recv=  5133 resp=5133 POST sessionReplays/blob 623B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  6232 resp=6233 POST sessionReplays/meta 1203B status=200 rows=[('ended', 2)]
    server recv=  6294 resp=6294 POST sessionReplays/meta 1064B status=200 rows=[('active', 3)]
    server recv= 11268 resp=11268 POST sessionEvents 494B status=200 rows=1 seqs=[0]
    server recv= 11294 resp=11295 POST sessionReplays/blob 623B status=200 chunkSeq=1 checkpoint=1 events=2

### BUG-5: ~60 kB of spans pending at reload; meta(ended) + events rejected
    file: evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
    marks: goto@27, ready@137, first replay blob received (or replay disabled)@5140, payload created@5359, reload-start@5659, reload-load-event@5680, reload-ready@5703
    console t=  5676 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    console t=  5676 warning [MapleClientSDK] traces flush skipped (cooldown 59995ms remaining)
    journal t=    83 script-start {"docId": "moitk4", "variant": "flush", "origin": "same", "autoFlushInterval": 60000, "replayEnabled": true}
    journal t=   109 runtime-ready {}
    journal t=  5358 spans-created {"n": 150, "attrBytes": 0, "prefix": "E-span"}
    journal t=  5663 pagehide(page listener, registered before SDK) {"docId": "moitk4", "persisted": false}
    journal t=  5668 visibilitychange:hidden {"docId": "moitk4"}
    journal t=  5669 console.error {"docId": "moitk4", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch"}
    journal t=  5674 console.warn {"docId": "moitk4", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59995ms remaining) "}
    journal t=  5700 script-start {"docId": "kknzhk", "variant": "flush", "origin": "same", "autoFlushInterval": 60000, "replayEnabled": true}
    journal t=  5702 runtime-ready {}
    playwright t=   136 request  POST sessionReplays/meta body=1058B
    playwright t=   188 finished POST sessionReplays/meta status=200
    playwright t=  5088 request  POST sessionEvents body=488B
    playwright t=  5137 request  POST sessionReplays/blob body=620B
    playwright t=  5139 finished POST sessionEvents status=200
    playwright t=  5189 finished POST sessionReplays/blob status=200
    playwright t=  5675 request  POST traces body=60652B
    playwright t=  5675 request  POST sessionReplays/meta body=6377B
    playwright t=  5676 request  POST sessionEvents body=21041B
    playwright t=  5676 failed   POST traces [net::ERR_ABORTED]
    playwright t=  5726 request  POST sessionReplays/meta body=1058B
    playwright t=  5777 finished POST sessionReplays/meta status=200
    playwright t= 10703 request  POST sessionEvents body=488B
    playwright t= 10727 request  POST sessionReplays/blob body=620B
    playwright t= 10754 finished POST sessionEvents status=200
    playwright t= 10778 finished POST sessionReplays/blob status=200
    server recv=   136 resp=136 POST sessionReplays/meta 1058B status=200 rows=[('active', 1)]
    server recv=  5088 resp=5088 POST sessionEvents 488B status=200 rows=1 seqs=[0]
    server recv=  5137 resp=5137 POST sessionReplays/blob 620B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  5667 resp=5668 POST traces 60652B status=200 spans=150 ['E-span-0', 'E-span-1', 'E-span-2', 'E-span-3']...
    server recv=  5726 resp=5726 POST sessionReplays/meta 1058B status=200 rows=[('active', 3)]
    server recv= 10702 resp=10702 POST sessionEvents 488B status=200 rows=1 seqs=[0]
    server recv= 10726 resp=10726 POST sessionReplays/blob 620B status=200 chunkSeq=1 checkpoint=1 events=2

### BUG-5: same with replay disabled; session rows go first, traces rejected
    file: evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
    marks: goto@26, ready@109, first replay blob received (or replay disabled)@109, payload created@331, reload-start@632, reload-load-event@650, reload-ready@675
    console t=   646 warning [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    journal t=    76 script-start {"docId": "x881i5", "variant": "flush", "origin": "same", "autoFlushInterval": 60000, "replayEnabled": false}
    journal t=   105 runtime-ready {}
    journal t=   330 spans-created {"n": 150, "attrBytes": 0, "prefix": "E-span"}
    journal t=   636 pagehide(page listener, registered before SDK) {"docId": "x881i5", "persisted": false}
    journal t=   640 visibilitychange:hidden {"docId": "x881i5"}
    journal t=   641 console.warn {"docId": "x881i5", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Fai
    journal t=   669 script-start {"docId": "pep5s3", "variant": "flush", "origin": "same", "autoFlushInterval": 60000, "replayEnabled": false}
    journal t=   671 runtime-ready {}
    playwright t=   106 request  POST sessionReplays/meta body=1068B
    playwright t=   159 finished POST sessionReplays/meta status=200
    playwright t=   642 request  POST sessionReplays/meta body=6386B
    playwright t=   642 request  POST sessionEvents body=497B
    playwright t=   646 request  POST traces body=60652B
    playwright t=   646 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=   646 failed   POST sessionEvents [net::ERR_ABORTED]
    playwright t=   672 request  POST sessionReplays/meta body=1068B
    playwright t=   724 finished POST sessionReplays/meta status=200
    server recv=    83 resp=83 POST sessionReplays/meta 1068B status=200 rows=[('active', 1)]
    server recv=   637 resp=638 POST sessionReplays/meta 6386B status=200 rows=[('ended', 2)]
    server recv=   638 resp=638 POST sessionEvents 497B status=200 rows=1 seqs=[0]
    server recv=   671 resp=671 POST sessionReplays/meta 1068B status=200 rows=[('active', 3)]

### BUG-6: replay chunk seq 1 consumed at unload, never POSTed (server sees 0 then 2)
    file: evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
    marks: goto@26, ready@125, first replay blob received (or replay disabled)@5135, payload created@5344, reload-start@5644, reload-load-event@5663, reload-ready@5689
    console t=  5656 warning [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    console t=  5656 warning [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    playwright t=   125 request  POST sessionReplays/meta body=1058B
    playwright t=   176 finished POST sessionReplays/meta status=200
    playwright t=  5080 request  POST sessionEvents body=488B
    playwright t=  5125 request  POST sessionReplays/blob body=620B
    playwright t=  5132 finished POST sessionEvents status=200
    playwright t=  5177 finished POST sessionReplays/blob status=200
    playwright t=  5655 request  POST sessionReplays/meta body=1128B
    playwright t=  5656 request  POST sessionEvents body=2100B
    playwright t=  5656 failed   POST sessionEvents [net::ERR_ABORTED]
    playwright t=  5656 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  5713 request  POST sessionReplays/meta body=1058B
    playwright t=  5765 finished POST sessionReplays/meta status=200
    playwright t= 10688 request  POST sessionEvents body=488B
    playwright t= 10714 request  POST sessionReplays/blob body=619B
    playwright t= 10740 finished POST sessionEvents status=200
    playwright t= 10765 finished POST sessionReplays/blob status=200
    server recv=   124 resp=125 POST sessionReplays/meta 1058B status=200 rows=[('active', 1)]
    server recv=  5080 resp=5080 POST sessionEvents 488B status=200 rows=1 seqs=[0]
    server recv=  5125 resp=5125 POST sessionReplays/blob 620B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  5652 resp=5652 POST sessionReplays/meta 1128B status=200 rows=[('ended', 2)]
    server recv=  5653 resp=5653 POST sessionEvents 2100B status=200 rows=3 seqs=[1, 2, 3]
    server recv=  5713 resp=5713 POST sessionReplays/meta 1058B status=200 rows=[('active', 3)]
    server recv= 10688 resp=10688 POST sessionEvents 488B status=200 rows=1 seqs=[0]
    server recv= 10713 resp=10713 POST sessionReplays/blob 619B status=200 chunkSeq=2 checkpoint=1 events=2

### BUG-7: span + log, reload 150 ms later: delivered, yet 'failed' in the old document
    file: evidence/K-reload-150ms-after-action/replay-on(default)-run0.json
    marks: goto@28, ready@136, action (span + log) done@1644, reload-start@1795, reload-load-event@1815, reload-ready@1842
    console t=  1808 warning [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    console t=  1808 warning [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    console t=  1808 error   [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    console t=  1808 error   [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    console t=  1808 warning [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)
    console t=  1808 warning [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)
    journal t=    84 script-start {"docId": "5eo9a3", "variant": "flush", "origin": "same", "autoFlushInterval": 1000, "replayEnabled": true}
    journal t=   108 runtime-ready {}
    journal t=  1640 span-created {"name": "K-action-span"}
    journal t=  1644 logs-created {"n": 1, "bytes": 0, "prefix": "K-action-log"}
    journal t=  1799 pagehide(page listener, registered before SDK) {"docId": "5eo9a3", "persisted": false}
    journal t=  1804 visibilitychange:hidden {"docId": "5eo9a3"}
    journal t=  1806 console.warn {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: F
    journal t=  1806 console.warn {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Fai
    journal t=  1807 console.error {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch"}
    journal t=  1807 console.error {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch"}
    journal t=  1808 console.warn {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] traces flush skipped (cooldown 59999ms remaining) "}
    journal t=  1808 console.warn {"docId": "5eo9a3", "afterPagehide": true, "visibility": "hidden", "text": "[MapleClientSDK] logs flush skipped (cooldown 59999ms remaining) "}
    journal t=  1838 script-start {"docId": "spoowt", "variant": "flush", "origin": "same", "autoFlushInterval": 1000, "replayEnabled": true}
    journal t=  1840 runtime-ready {}
    playwright t=   136 request  POST sessionReplays/meta body=1060B
    playwright t=   187 finished POST sessionReplays/meta status=200
    playwright t=  1807 request  POST traces body=1165B
    playwright t=  1807 request  POST logs body=1014B
    playwright t=  1807 request  POST sessionReplays/meta body=1164B
    playwright t=  1807 request  POST sessionEvents body=1021B
    playwright t=  1807 failed   POST logs [net::ERR_ABORTED]
    playwright t=  1807 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  1807 failed   POST sessionEvents [net::ERR_ABORTED]
    playwright t=  1807 failed   POST traces [net::ERR_ABORTED]
    playwright t=  1871 request  POST sessionReplays/meta body=1060B
    playwright t=  1922 finished POST sessionReplays/meta status=200
    server recv=   136 resp=136 POST sessionReplays/meta 1060B status=200 rows=[('active', 1)]
    server recv=  1801 resp=1801 POST traces 1165B status=200 spans=1 ['K-action-span']
    server recv=  1802 resp=1802 POST logs 1014B status=200 logs=1 ['K-action-log-0 ']
    server recv=  1802 resp=1802 POST sessionReplays/meta 1164B status=200 rows=[('ended', 2)]
    server recv=  1804 resp=1804 POST sessionEvents 1021B status=200 rows=2 seqs=[0, 1]
    server recv=  1870 resp=1871 POST sessionReplays/meta 1060B status=200 rows=[('active', 3)]

### BUG-8: periodic (non-keepalive) sessionEvents POST in flight at reload
    file: evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run0.json
    marks: goto@29, ready@140, periodic sessionEvents POST received by server; response held 3 s@5095, reload-start@5195, reload-load-event@5209, reload-ready@5243
    console t=  5204 warning [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    console t=  5204 warning [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    journal t=    82 script-start {"docId": "gndnpy", "variant": "flush", "origin": "same", "replayEnabled": true}
    journal t=   108 runtime-ready {}
    journal t=  5199 pagehide(page listener, registered before SDK) {"docId": "gndnpy", "persisted": false}
    journal t=  5201 visibilitychange:hidden {"docId": "gndnpy"}
    journal t=  5202 console.warn {"docId": "gndnpy", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: F
    journal t=  5202 console.warn {"docId": "gndnpy", "afterPagehide": true, "visibility": "hidden", "text": "[maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Fai
    journal t=  5231 script-start {"docId": "qm3r0p", "variant": "flush", "origin": "same", "replayEnabled": true}
    journal t=  5233 runtime-ready {}
    playwright t=   139 request  POST sessionReplays/meta body=1054B
    playwright t=   190 finished POST sessionReplays/meta status=200
    playwright t=  5089 request  POST sessionEvents body=2959B
    playwright t=  5139 request  POST sessionReplays/blob body=619B
    playwright t=  5191 finished POST sessionReplays/blob status=200
    playwright t=  5204 request  POST sessionReplays/meta body=1124B
    playwright t=  5204 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  5204 failed   POST sessionEvents [net::ERR_ABORTED]
    playwright t=  5258 request  POST sessionReplays/meta body=1054B
    playwright t=  5310 finished POST sessionReplays/meta status=200
    server recv=   138 resp=138 POST sessionReplays/meta 1054B status=200 rows=[('active', 1)]
    server recv=  5088 resp=8089 POST sessionEvents 2959B status=200 rows=6 seqs=[0, 1, 2, 3, 4, 5]
    server recv=  5139 resp=5139 POST sessionReplays/blob 619B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  5201 resp=5201 POST sessionReplays/meta 1124B status=200 rows=[('ended', 2)]
    server recv=  5258 resp=5258 POST sessionReplays/meta 1054B status=200 rows=[('active', 3)]

### BUG-9: Maple.layer, H-tail lost
    file: evidence/H-maple-layer-unload/layer-run0.json
    marks: goto@28, ready@117, H-early created@120, H-tail created@6623, reload-start@7623, reload-load-event@7635, reload-ready@7662
    playwright t=   153 request  POST sessionReplays/meta body=1039B
    playwright t=   204 finished POST sessionReplays/meta status=200
    playwright t=  5112 request  POST sessionEvents body=469B
    playwright t=  5118 request  POST traces body=1066B
    playwright t=  5153 request  POST sessionReplays/blob body=603B
    playwright t=  5164 finished POST sessionEvents status=200
    playwright t=  5169 finished POST traces status=200
    playwright t=  5205 finished POST sessionReplays/blob status=200
    playwright t=  7631 request  POST sessionReplays/meta body=1178B
    playwright t=  7631 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  7686 request  POST sessionReplays/meta body=1039B
    playwright t=  7737 finished POST sessionReplays/meta status=200
    server recv=   152 resp=152 POST sessionReplays/meta 1039B status=200 rows=[('active', 1)]
    server recv=  5112 resp=5112 POST sessionEvents 469B status=200 rows=1 seqs=[0]
    server recv=  5118 resp=5118 POST traces 1066B status=200 spans=1 ['H-early']
    server recv=  5153 resp=5153 POST sessionReplays/blob 603B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  7629 resp=7629 POST sessionReplays/meta 1178B status=200 rows=[('ended', 2)]
    server recv=  7685 resp=7685 POST sessionReplays/meta 1039B status=200 rows=[('active', 3)]

### Artifact: Playwright requestfailed vs server 200
    file: evidence/G-playwright-keepalive-artifact/with-final-reload-run0.json
    marks: goto@30, ready@112, reload-start@8309, reload-load-event@8325, reload-ready@8347
    playwright t=   140 request  POST sessionReplays/meta body=1064B
    playwright t=   191 finished POST sessionReplays/meta status=200
    playwright t=  1110 request  POST traces body=1941B
    playwright t=  1163 finished POST traces status=200
    playwright t=  2109 request  POST traces body=1548B
    playwright t=  2162 finished POST traces status=200
    playwright t=  3110 request  POST traces body=1941B
    playwright t=  3161 finished POST traces status=200
    playwright t=  4109 request  POST traces body=1548B
    playwright t=  4161 finished POST traces status=200
    playwright t=  5091 request  POST sessionEvents body=494B
    playwright t=  5109 request  POST traces body=1944B
    playwright t=  5140 request  POST sessionReplays/blob body=620B
    playwright t=  5142 finished POST sessionEvents status=200
    playwright t=  5161 finished POST traces status=200
    playwright t=  5192 finished POST sessionReplays/blob status=200
    playwright t=  6110 request  POST traces body=1550B
    playwright t=  6161 finished POST traces status=200
    playwright t=  7110 request  POST traces body=1944B
    playwright t=  7162 finished POST traces status=200
    playwright t=  8110 request  POST traces body=1550B
    playwright t=  8161 finished POST traces status=200
    playwright t=  8319 request  POST traces body=1158B
    playwright t=  8319 request  POST sessionReplays/meta body=1868B
    playwright t=  8319 failed   POST sessionReplays/meta [net::ERR_ABORTED]
    playwright t=  8319 failed   POST traces [net::ERR_ABORTED]
    playwright t=  8370 request  POST sessionReplays/meta body=1064B
    playwright t=  8422 finished POST sessionReplays/meta status=200
    server recv=   139 resp=139 POST sessionReplays/meta 1064B status=200 rows=[('active', 1)]
    server recv=  1111 resp=1111 POST traces 1941B status=200 spans=3 ['G-0', 'G-1', 'G-2']
    server recv=  2109 resp=2109 POST traces 1548B status=200 spans=2 ['G-3', 'G-4']
    server recv=  3109 resp=3109 POST traces 1941B status=200 spans=3 ['G-5', 'G-6', 'G-7']
    server recv=  4109 resp=4109 POST traces 1548B status=200 spans=2 ['G-8', 'G-9']
    server recv=  5090 resp=5091 POST sessionEvents 494B status=200 rows=1 seqs=[0]
    server recv=  5109 resp=5109 POST traces 1944B status=200 spans=3 ['G-10', 'G-11', 'G-12']
    server recv=  5140 resp=5140 POST sessionReplays/blob 620B status=200 chunkSeq=0 checkpoint=1 events=2
    server recv=  6109 resp=6109 POST traces 1550B status=200 spans=2 ['G-13', 'G-14']
    server recv=  7109 resp=7109 POST traces 1944B status=200 spans=3 ['G-15', 'G-16', 'G-17']
    server recv=  8109 resp=8109 POST traces 1550B status=200 spans=2 ['G-18', 'G-19']
    server recv=  8316 resp=8316 POST traces 1158B status=200 spans=1 ['G-tail']
    server recv=  8317 resp=8317 POST sessionReplays/meta 1868B status=200 rows=[('ended', 2)]
    server recv=  8370 resp=8370 POST sessionReplays/meta 1064B status=200 rows=[('active', 3)]


==============================================================================
APPENDIX C. FILES IN THIS FOLDER
==============================================================================

MAPLE_BUG_REPORT.txt   -- this report
evidence/   -- raw outputs of the runs cited in this report
  environment-shell-A.json
  environment-shell-B_A0_C_D_Dr_E_F_G_H_I_K_L.json
  environment-shell-W.json
  runner-console-A.txt
  runner-console-F-rerun.txt
  runner-console-W.txt
  runner-console-main.txt
  server-requests-all.jsonl
  summary-shell-2026-10-07T23-39-09-001Z.txt
  summary-shell-2026-10-07T23-39-09-004Z.txt
  summary-shell-2026-10-07T23-58-29-444Z.txt
  summary.jsonl
  webkit-firefox-launch.txt
  A-oversized-batch/
    (8 run files: <variant>-run<N>.json)
  A0-threshold/
    (8 run files: <variant>-run<N>.json)
  B-keepalive-probe/
    (2 run files: <variant>-run<N>.json)
  B-sdk-concurrent-signals/
    (6 run files: <variant>-run<N>.json)
  C-unload-during-cooldown/
    (12 run files: <variant>-run<N>.json)
  D-hang-head-of-line-blocking/
    (2 run files: <variant>-run<N>.json)
  D-pagehide-behind-inflight-flush/
    (12 run files: <variant>-run<N>.json)
  Dr-random-reload-latency/
    (40 run files: <variant>-run<N>.json)
  E-combined-unload-budget/
    (15 run files: <variant>-run<N>.json)
  F-replay-posts-aborted-by-navigation/
    (8 run files: <variant>-run<N>.json)
  Fr-random-reloads/
    (4 run files: <variant>-run<N>.json)
  G-playwright-keepalive-artifact/
    (4 run files: <variant>-run<N>.json)
  H-maple-layer-unload/
    (9 run files: <variant>-run<N>.json)
  I-cross-origin/
    (4 run files: <variant>-run<N>.json)
  K-reload-150ms-after-action/
    (6 run files: <variant>-run<N>.json)
  L-multi-arg-log-body/
    (1 run files: <variant>-run<N>.json)
  W-workaround-keepalive-downgrade-shim/
    (4 run files: <variant>-run<N>.json)
  headless-new/
    environment-new-B_G_K_E.json
    runner-console.txt
    server-requests-all.jsonl
    summary-new-2026-10-07T23-50-02-862Z.txt
    summary.jsonl
    B-keepalive-probe/
      (2 run files: <variant>-run<N>.json)
    B-sdk-concurrent-signals/
      (4 run files: <variant>-run<N>.json)
    E-combined-unload-budget/
      (10 run files: <variant>-run<N>.json)
    G-playwright-keepalive-artifact/
      (4 run files: <variant>-run<N>.json)
    K-reload-150ms-after-action/
      (4 run files: <variant>-run<N>.json)
  sdk-0.10.0/
    environment-shell-sdk0.10.0-A.json
    environment-shell-sdk0.10.0-A0_B_C_D_E_K.json
    runner-console-A.txt
    runner-console-main.txt
    server-requests-all.jsonl
    summary-shell-sdk0.10.0-2026-10-07T23-41-29-113Z.txt
    summary-shell-sdk0.10.0-2026-10-07T23-43-48-971Z.txt
    summary.jsonl
    A-oversized-batch/
      (8 run files: <variant>-run<N>.json)
    A0-threshold/
      (16 run files: <variant>-run<N>.json)
    B-keepalive-probe/
      (2 run files: <variant>-run<N>.json)
    B-sdk-concurrent-signals/
      (4 run files: <variant>-run<N>.json)
    C-unload-during-cooldown/
      (8 run files: <variant>-run<N>.json)
    D-hang-head-of-line-blocking/
      (2 run files: <variant>-run<N>.json)
    D-pagehide-behind-inflight-flush/
      (8 run files: <variant>-run<N>.json)
    E-combined-unload-budget/
      (10 run files: <variant>-run<N>.json)
    K-reload-150ms-after-action/
      (4 run files: <variant>-run<N>.json)
repro/   -- standalone Bun + Playwright reproduction project (see repro/README.md)
  .gitignore
  README.md   -- how to run; what each scenario does and what to expect
  build.ts   -- Bun.build of the browser bundles (+ optional 0.10.0 variant)
  bun.lock   -- lockfile
  lib.ts   -- Playwright harness (capture console, network events, server records, evidence writing)
  package.json   -- pins @maple-dev/effect-sdk 0.8.2, effect 4.0.0-rc.112, playwright 1.57.0
  run-scenarios.ts   -- scenario runner + summary
  server.ts   -- mock Maple/OTLP ingest + static server; request logging; latency/failure injection
  client/
    app.ts   -- test page using MapleFlush.make / Maple.layer (window.T API)
    probe.ts   -- SDK-free keepalive probe (window.P API)
  public/
    index.html   -- test page shell
    probe.html   -- probe page shell
  scenarios/
    b-probe.ts   -- scenario B part 1: Chromium keepalive limits without the SDK
    sdk.ts   -- all SDK scenarios (A0, A, Bsdk, C, D, Dhang, Dr, E, F, Fr, G, H, I, K, L, W)
  variants/
    sdk-0.10.0/
      bun.lock
      package.json   -- optional: same page against @maple-dev/effect-sdk 0.10.0 + effect 4.0.0
sdk-excerpts/   -- verbatim SDK (and Effect) source excerpts with file names + line numbers
  01-flush-core-cooldown-serialization.txt
  02-buffers-cap-restore.txt
  03-mapleflush-make-client.txt
  04-replay-transport-keepalive-budget.txt
  05-session-event-sink.txt
  06-session-lifecycle-unload.txt
  07-client-session-startup.txt
  08-replay-record-final-chunk.txt
  09-maple-layer-client.txt
  10-effect-otlp-exporter.txt
  11-readme.txt
  12-v0.10.0-what-changed-and-what-did-not.txt

Evidence conventions:
  evidence/<scenario>/<variant>-run<N>.json   one run: config, result, marks (timeline), full browser console,
                                              Playwright request/requestfinished/requestfailed for ingest URLs,
                                              every server record of that run (t* = ms since run start), and the
                                              page's localStorage journal (lifecycle + SDK console lines tagged per
                                              document)
  evidence/server-requests-all.jsonl          every ingest request of the 0.8.2 runs (absolute epoch ms)
  evidence/summary.jsonl                      per-run verdict lines (authoritative; Appendix A is generated from it)
  evidence/summary-*.txt                      snapshot written at the end of each runner invocation (the main one
                                              predates the F periodic-events re-run and scenario W: see
                                              runner-console-F-rerun.txt and runner-console-W.txt)
  evidence/runner-console-*.txt               raw stdout of the runner (includes full stack traces)
  evidence/environment-*.json                 browser/Playwright/Bun versions + UA per runner invocation
  evidence/sdk-0.10.0/                        same layout, runs against @maple-dev/effect-sdk 0.10.0 + effect 4.0.0
  evidence/headless-new/                      same layout, SDK 0.8.2 in full Chromium "new headless"
  evidence/webkit-firefox-launch.txt          why WebKit/Firefox were not tested

@@@@@@@@@@ END FILE: MAPLE_BUG_REPORT.txt

@@@@@@@@@@ FILE: repro/README.md (8750 bytes)
# Repro: @maple-dev/effect-sdk 0.8.2 browser flush / keepalive data loss

A standalone project (no dependency on anything outside this folder) that reproduces the
issues described in `../MAPLE_BUG_REPORT.txt` in headless Chromium, against a mock
Maple/OTLP ingest that records every request it receives.

Pinned versions: `@maple-dev/effect-sdk@0.8.2`, `effect@4.0.0-rc.112` (the SDK's Effect
peer), `playwright@1.57.0` (Chromium build 1200 = Chromium 143.0.7499.4). Bun 1.4.2 was used.

## Quick start

```sh
cd repro
bun install
bunx playwright install chromium   # only if Chromium build 1200 is not already in ~/.cache/ms-playwright

bun run start        # builds the bundles and serves http://localhost:4173/ (manual poking)
bun run scenarios    # builds, then runs every scenario in headless Chromium (~40 min; group A alone ~2.5 min in parallel browsers)
```

Run a subset / change settings:

```sh
bun run build.ts && bun run run-scenarios.ts B C D        # only groups B, C and D
REPEAT=1 bun run run-scenarios.ts E                        # 1 run per variant instead of the default
HEADLESS=new bun run run-scenarios.ts G                    # full Chromium "new headless" (channel "chromium")
EVIDENCE_DIR=../evidence bun run run-scenarios.ts          # where evidence JSON goes (default ./evidence-out)
PAGE_PORT=4183 CROSS_PORT=4184 bun run run-scenarios.ts A  # other ports (lets two runners run side by side)
```

Optional — the same page against the newest published SDK (0.10.0, Effect 4.0.0):

```sh
(cd variants/sdk-0.10.0 && bun install)
bun run build.ts                                   # now also emits public/dist-0.10.0/
SDK=0.10.0 bun run run-scenarios.ts A0 B C D E K   # or open http://localhost:4173/?sdk=0.10.0&tag=x
```

## Layout

| file | what |
| --- | --- |
| `server.ts` | Bun server. Port 4173: test pages + same-origin ingest under `/otlp/<tag>/v1/*` + `/probe/sink`. Port 4174: the same ingest cross-origin (CORS, preflight with `access-control-max-age: 0`). Logs every request (timestamps for headers-received / body-read / response-sent, byte size, headers of interest, parsed span names / log bodies / metric names / session-event types+seqs / replay chunk headers) to memory and to a JSONL file. Latency / failure injection via rules (`latencyMs`, `latencyCount`, `failStatus`, `failCount`, scoped per run tag). Control API: `POST /__control {rules}`, `GET /__log?tag=`, `POST /__reset`. |
| `client/app.ts` | Test page. Query string: `tag`, `variant=flush` (MapleFlush.make) / `layer` (Maple.layer) / `layer-flusher` (Maple.layer + user-land pagehide flush via Effect's `OtlpExporter.Flusher` and `FetchHttpClient.RequestInit {keepalive:true}`), `origin=same/cross`, `interval=<ms>/false` (autoFlushInterval), `replay=1/0`. Uses `ingestKey: "~"` like a same-origin proxy that injects the real key server-side. Exposes `window.T` (`spans(n, attrBytes, prefix)`, `span(name)`, `logs(n, bytes, prefix)`, `logMulti()`, `metric()`, `flush()`, `mutate(chars, chunks)`, `journal()`). Journals lifecycle events and every `[MapleClientSDK]`/`[maple]` console warning/error to localStorage tagged with a per-document id, so messages printed during unload can be attributed to the document that printed them. |
| `client/probe.ts` + `public/probe.html` | SDK-free page for measuring Chromium's `fetch(keepalive)` limits (`window.P.send(bytes, {delay, keepalive, cross, contentType, consume, utf8TwoByte})`, `P.armUnloadBurst(sizes)`). |
| `build.ts` | `Bun.build` (target browser, ESM, code splitting so the SDK's replay engine is a lazy chunk, like in a real app). |
| `lib.ts` | Harness: launches Chromium, captures browser console + Playwright `request`/`requestfinished`/`requestfailed`, joins with server records, writes evidence JSON per run. |
| `scenarios/b-probe.ts` | Scenario B part 1 (keepalive limits, no SDK). |
| `scenarios/sdk.ts` | All SDK scenarios. |
| `run-scenarios.ts` | Runs groups, prints + writes a summary. |

Every run writes `<EVIDENCE_DIR>/<scenario>/<variant>-run<N>.json` with: config, result,
marks (timeline), the full browser console, every Playwright network event for ingest URLs,
every server record for that run (all times in ms relative to run start), and the page's
localStorage journal. `<EVIDENCE_DIR>/server-requests-all.jsonl` has every ingest request
(absolute epoch ms), `summary-*.txt` the aggregated verdicts.

## Scenarios (groups) and what to expect with 0.8.2

| group | scenario | expected / observed with 0.8.2 |
| --- | --- | --- |
| B | `B-keepalive-probe` (no SDK) | Chromium rejects a keepalive request immediately with `TypeError: Failed to fetch` when the *sum* of in-flight keepalive bodies of the document would exceed 65,536 bytes (UTF-8 bytes). Per document (other tabs and the next document after navigation have their own budget). Released when the response body has finished loading. Cross-origin keepalive + preflight works. |
| B | `B-sdk-concurrent-signals` | 40 spans (≈38 KB) + 30 logs (≈32 KB): each fits, but traces/logs/metrics are POSTed concurrently, so the logs POST is rejected → `[MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch`. On reload the logs (and the replay events batch) are lost. |
| A0 | `A0-threshold` | ≈ 396 B per plain span → a single flush of ≥ 164 spans (> 65,536 B body) is rejected. |
| A | `A-oversized-batch` | 200 spans × 500 B in one flush interval: the batch is rejected, restored, retried after 60 s with even more spans, rejected again… **no span is ever delivered for the rest of the page's life**, while logs keep flowing. |
| C | `C-unload-during-cooldown` | One 503 from ingest → 60 s cooldown → the pagehide/visibilitychange flush is skipped (`traces flush skipped (cooldown …)`) → the failed batch and every span created since are lost on reload / navigation / close. |
| D | `D-pagehide-behind-inflight-flush`, `D-hang-head-of-line-blocking` | If one traces POST is slow (held 4–7 s by the mock), the pagehide flush is queued behind it by `makeSerializedFlush` and never runs before the document dies → tail spans lost (also a span that ended 5 s before the reload). One slow POST also delays *all* later traces **and logs** exports. |
| Dr | `Dr-random-reload-latency` | Same as D, probabilistic: steady spans, reload at a random time, 150–600 ms ingest latency. Counts spans lost. |
| E | `E-combined-unload-budget` | At unload the MapleFlush OTLP posts and the replay/session posts share one 64 KiB budget; whichever listener fires first wins. ≈ 60 KB of spans → the session `ended` metadata row and final events batch are rejected; with replay disabled the metadata-only session fires first and the *traces* POST is rejected. Separately, the final replay chunk is never posted on unload at all (its chunk sequence number is consumed, gzip is async) — also with `Maple.layer`. |
| F | `F-replay-posts-aborted-by-navigation`, `Fr-random-reloads` | Non-keepalive session posts (5 s events flush, `active` metadata row) in flight during a navigation are aborted → `[maple] session replay … POST failed (dropping; will retry on next chunk)`; nothing is retried. |
| G | `G-playwright-keepalive-artifact` | Steady state: Playwright reports keepalive requests as finished. Any request still in flight when the document unloads is reported as `requestfailed net::ERR_ABORTED` even though the server answered 200. |
| H | `H-maple-layer-unload` | `Maple.layer`: span created 1 s before reload is lost (documented). `layer-flusher` workaround delivers it. MapleFlush delivers it. |
| I | `I-cross-origin` | Cross-origin endpoint works (preflight requests `authorization,content-type,x-maple-sdk`; Chromium ignores the SDK's `user-agent` header). |
| K | `K-reload-150ms-after-action` | 1 span + 1 Effect log, reload 150 ms later: both are delivered (server 200), yet the *old* document prints `traces flush failed`, `logs flush failed`, `traces/logs flush skipped (cooldown 60000ms remaining)`, `[maple] session replay events/metadata POST failed` — false alarms caused by the document being torn down while its keepalive requests complete. |
| W | `W-workaround-keepalive-downgrade-shim` | A-style burst (200 × 500 B) without / with a user-land `window.fetch` wrapper (`?shim=1`, see `client/app.ts`) that sends keepalive POSTs above ~60 KiB as normal requests. Without: 0/200 burst spans and 0/5 later spans delivered. With: 200/200 and 5/5. |
| L | `L-multi-arg-log-body` | `Effect.logError(a, b, c, obj)` is exported as an OTLP `arrayValue` body (structured), not a JSON string. |

WebKit could not be launched on the machine used for the report (missing system
libraries), and Firefox was not available, so only Chromium results exist.
@@@@@@@@@@ END FILE: repro/README.md

@@@@@@@@@@ FILE: repro/package.json (379 bytes)
{
  "name": "maple-effect-sdk-keepalive-repro",
  "private": true,
  "type": "module",
  "scripts": {
    "build": "bun run build.ts",
    "start": "bun run build.ts && bun run server.ts",
    "scenarios": "bun run build.ts && bun run run-scenarios.ts"
  },
  "dependencies": {
    "@maple-dev/effect-sdk": "0.8.2",
    "effect": "4.0.0-rc.112",
    "playwright": "1.57.0"
  }
}
@@@@@@@@@@ END FILE: repro/package.json

@@@@@@@@@@ FILE: repro/.gitignore (153 bytes)
node_modules/
public/dist/
variants/sdk-0.10.0/app.ts
variants/sdk-0.10.0/node_modules/
public/dist-0.10.0/
scratch/
server-requests.jsonl
evidence-out/
@@@@@@@@@@ END FILE: repro/.gitignore

@@@@@@@@@@ FILE: repro/build.ts (1913 bytes)
// Bundle the browser entry points with Bun (code splitting on, so the SDK's
// dynamic import of the replay engine becomes its own chunk, as in a real app).
//
// Main build:     client/app.ts against @maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112 -> public/dist/
// Optional build: if variants/sdk-0.10.0/node_modules exists (cd variants/sdk-0.10.0 && bun install),
//                 the same app.ts is also built against @maple-dev/effect-sdk@0.10.0 + effect@4.0.0
//                 -> public/dist-0.10.0/ (load the page with ?sdk=0.10.0). Effect 4.0.0 renamed
//                 effect/unstable/{http,observability} to effect/{http,observability}; the copy is
//                 rewritten accordingly. Nothing else differs.
import { existsSync, readFileSync, writeFileSync } from "node:fs"

const build = async (entrypoints: string[], outdir: string) => {
  const result = await Bun.build({
    entrypoints,
    outdir,
    target: "browser",
    format: "esm",
    splitting: true,
    minify: false,
    naming: { entry: "[name].js", chunk: "chunk-[hash].js" },
  })
  if (!result.success) {
    for (const log of result.logs) console.error(log)
    process.exit(1)
  }
  for (const o of result.outputs) console.log(`built ${o.path.replace(process.cwd() + "/", "")} (${o.size} B)`)
}

await build(["./client/app.ts", "./client/probe.ts"], "./public/dist")

if (existsSync("./variants/sdk-0.10.0/node_modules/@maple-dev/effect-sdk")) {
  const src = readFileSync("./client/app.ts", "utf8").replaceAll("effect/unstable/http", "effect/http").replaceAll("effect/unstable/observability", "effect/observability")
  writeFileSync("./variants/sdk-0.10.0/app.ts", `// GENERATED by build.ts from client/app.ts — do not edit\n${src}`)
  await build(["./variants/sdk-0.10.0/app.ts"], "./public/dist-0.10.0")
} else {
  console.log("(variants/sdk-0.10.0 not installed — skipping the optional 0.10.0 build)")
}

@@@@@@@@@@ END FILE: repro/build.ts

@@@@@@@@@@ FILE: repro/bun.lock (4821 bytes)
{
  "lockfileVersion": 2,
  "configVersion": 1,
  "workspaces": {
    "": {
      "name": "maple-effect-sdk-keepalive-repro",
      "dependencies": {
        "@maple-dev/effect-sdk": "0.8.2",
        "effect": "4.0.0-rc.112",
        "playwright": "1.57.0",
      },
    },
  },
  "packages": {
    "@maple-dev/effect-sdk": ["@maple-dev/effect-sdk@0.8.2", "", { "dependencies": { "std-env": "^4.0.0" }, "peerDependencies": { "effect": ">=4.0.0-rc.111" } }, "sha512-bbkmSWI/RVcLWyFL1czayCw+wCCq3v0vYM3n08/O0rT63dwvHepYy+2cV0rXHo0FPMplvi5FUV0lKAQobJ0DvQ=="],

    "@msgpackr-extract/msgpackr-extract-darwin-arm64": ["@msgpackr-extract/msgpackr-extract-darwin-arm64@3.0.4", "", { "os": "darwin", "cpu": "arm64" }, "sha512-LCkGo6JDfaBhgST7UpPWgNgLINpcpabaHfyz5OBx75nUYxBsaEPxjnyNjWpeb/xBup/682QnBfRBy2/LvPutZQ=="],

    "@msgpackr-extract/msgpackr-extract-darwin-x64": ["@msgpackr-extract/msgpackr-extract-darwin-x64@3.0.4", "", { "os": "darwin", "cpu": "x64" }, "sha512-zExlW9zUJKZH/tOtVMttwjKa4Xm/3KcNjnE3dPN92uCktwavMxpgCA3MoJK/DOnTWsQgo224OaST27/mPNAf+w=="],

    "@msgpackr-extract/msgpackr-extract-linux-arm": ["@msgpackr-extract/msgpackr-extract-linux-arm@3.0.4", "", { "os": "linux", "cpu": "arm" }, "sha512-Tg3yX65f5GbtXLkrYEHE5oibZG9epyYWas7FogTTEJeDEF9JlXJzKgXaNhT3UXlTOeA+AfZpYZYZ0uPj7Cfquw=="],

    "@msgpackr-extract/msgpackr-extract-linux-arm64": ["@msgpackr-extract/msgpackr-extract-linux-arm64@3.0.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-dgX0P/9wGPJeHFBG+ZmhgE6bmtMt7NP5CRBGyyktpopdk/mW4POnrpQsSLtKI1dwpc+pPLuXHDh6vvskyQE/sw=="],

    "@msgpackr-extract/msgpackr-extract-linux-x64": ["@msgpackr-extract/msgpackr-extract-linux-x64@3.0.4", "", { "os": "linux", "cpu": "x64" }, "sha512-8TNXMEjJc3QEy7R/x1INhgiU+XakDAFUzBhaz7+Rbrs8NH5UQeHQxxmzsSBJGyV6I1jW79undiQm8tOI+D+8FQ=="],

    "@msgpackr-extract/msgpackr-extract-win32-x64": ["@msgpackr-extract/msgpackr-extract-win32-x64@3.0.4", "", { "os": "win32", "cpu": "x64" }, "sha512-CmCXPQrkbwExx3j946/PtHWHbYJiCRBRDl4BlkRQcJB/YOwQxJRTpoo7aTsortjgoJ1x7opzTSxn7C+ASSLVjQ=="],

    "detect-libc": ["detect-libc@2.1.2", "", {}, "sha512-Btj2BOOO83o3WyH59e8MgXsxEQVcarkUOpEYrubB0urwnN10yQ364rsiByU11nZlqWYZm05i/of7io4mzihBtQ=="],

    "effect": ["effect@4.0.0-rc.112", "", { "dependencies": { "fast-check": "^4.9.0", "msgpackr": "^2.0.5" } }, "sha512-wXxwuh1Ywnv4cPRM3Wfa0vDwuOHnZ1TsTgHJkG9XgzND6inhBH9n1vBxhg3iIXOia/OrpmvVmd3lrD4vq6bF3A=="],

    "fast-check": ["fast-check@4.10.2", "", { "dependencies": { "pure-rand": "^8.0.0" } }, "sha512-iK2f+YrcmoeGqk6fA0ea2bptcu/itMIm4NfEozq6N25+aG6h7s5HZbB/k1aV7b5w5sFLMCbbtRUsTVR+BgC3xw=="],

    "fsevents": ["fsevents@2.3.2", "", { "os": "darwin" }, "sha512-xiqMQR4xAeHTuB9uWm+fFRcIOgKBMiOBP+eXiyT7jsgVCq1bkVygt00oASowB7EdtpOHaaPgKt812P9ab+DDKA=="],

    "msgpackr": ["msgpackr@2.1.0", "", { "optionalDependencies": { "msgpackr-extract": "^3.0.4" } }, "sha512-p/pBCVO63CsvvpkomUnNNag6+n38rULuDA6HHe70o2gtC8ODI52foF/4ko2qQcp6OiErJXTmrZeXmsGGHsIQNQ=="],

    "msgpackr-extract": ["msgpackr-extract@3.0.4", "", { "dependencies": { "node-gyp-build-optional-packages": "5.2.2" }, "optionalDependencies": { "@msgpackr-extract/msgpackr-extract-darwin-arm64": "3.0.4", "@msgpackr-extract/msgpackr-extract-darwin-x64": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-arm": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-arm64": "3.0.4", "@msgpackr-extract/msgpackr-extract-linux-x64": "3.0.4", "@msgpackr-extract/msgpackr-extract-win32-x64": "3.0.4" }, "bin": { "download-msgpackr-prebuilds": "bin/download-prebuilds.js" } }, "sha512-4kmO/MdyUIkLIvTPr8VHLil4AtoKIoniWPIEk5+CDy0xnWC84azhSFmuJ7PxZdsYtiP5kEeQsORAVIeMgxT+Hw=="],

    "node-gyp-build-optional-packages": ["node-gyp-build-optional-packages@5.2.2", "", { "dependencies": { "detect-libc": "^2.0.1" }, "bin": { "node-gyp-build-optional-packages": "bin.js", "node-gyp-build-optional-packages-optional": "optional.js", "node-gyp-build-optional-packages-test": "build-test.js" } }, "sha512-s+w+rBWnpTMwSFbaE0UXsRlg7hU4FjekKU4eyAih5T8nJuNZT1nNsskXpxmeqSK9UzkBl6UgRlnKc8hz8IEqOw=="],

    "playwright": ["playwright@1.57.0", "", { "dependencies": { "playwright-core": "1.57.0" }, "optionalDependencies": { "fsevents": "2.3.2" }, "bin": { "playwright": "cli.js" } }, "sha512-ilYQj1s8sr2ppEJ2YVadYBN0Mb3mdo9J0wQ+UuDhzYqURwSoW4n1Xs5vs7ORwgDGmyEh33tRMeS8KhdkMoLXQw=="],

    "playwright-core": ["playwright-core@1.57.0", "", { "bin": { "playwright-core": "cli.js" } }, "sha512-agTcKlMw/mjBWOnD6kFZttAAGHgi/Nw0CZ2o6JqWSbMlI219lAFLZZCyqByTsvVAJq5XA5H8cA6PrvBRpBWEuQ=="],

    "pure-rand": ["pure-rand@8.4.2", "", {}, "sha512-vvuOGgcuPJAirlHvuQw1TrOiw7ptaIXXmIbNuiNOY6lNGJJH49PQ1Kj4nd783nPdQhQdicgOjVI2yI/9BD6/Ng=="],

    "std-env": ["std-env@4.3.0", "", {}, "sha512-OtU/EgQ1kIm5KwqQpBC6ZEMXrZRui11w8zgfTWp8cdO9B8OaPsbA8bTHO2P+HNo1VlUTGMVBwPhydu6poeXiag=="],
  }
}
@@@@@@@@@@ END FILE: repro/bun.lock

@@@@@@@@@@ FILE: repro/client/app.ts (9255 bytes)
/**
 * Test page client. Configured entirely through the query string:
 *
 *   ?tag=<run id>            required; becomes part of the ingest endpoint path
 *   &variant=flush|layer|layer-flusher
 *        flush         -> MapleFlush.make(...)               (the configuration under test)
 *        layer         -> Maple.layer(...)                   (baseline, documented "no unload flush")
 *        layer-flusher -> Maple.layer(...) + user-land pagehide hook calling Effect's
 *                         OtlpExporter.Flusher, with FetchHttpClient.RequestInit {keepalive:true}
 *                         (a possible workaround, NOT something the SDK does)
 *   &origin=same|cross       same: endpoint = <page origin>/otlp/<tag>   (like a same-origin ingest proxy)
 *                            cross: endpoint = http://127.0.0.1:<cross port>/otlp/<tag>
 *   &interval=<ms>|false     MapleFlush autoFlushInterval (default: SDK default 5000)
 *   &replay=1|0              replay.enabled (default 1 = SDK default)
 *   &crossPort=4174
 *
 * Exposes `window.T` for the Playwright runner (and for manual poking in DevTools).
 */
import { Maple, MapleFlush } from "@maple-dev/effect-sdk/client"
import { Effect, Layer, ManagedRuntime, Metric } from "effect"
import { FetchHttpClient } from "effect/unstable/http"
import { OtlpExporter } from "effect/unstable/observability"

const params = new URLSearchParams(location.search)
const tag = params.get("tag") ?? "manual"
const variant = params.get("variant") ?? "flush"
const origin = params.get("origin") ?? "same"
const crossPort = params.get("crossPort") ?? "4174"
const intervalParam = params.get("interval")
const autoFlushInterval = intervalParam === null ? undefined : intervalParam === "false" ? (false as const) : Number(intervalParam)
const replayEnabled = (params.get("replay") ?? "1") === "1"

const endpoint = origin === "cross" ? `http://127.0.0.1:${crossPort}/otlp/${tag}` : `${location.origin}/otlp/${tag}`

// --- lifecycle journal (survives reloads via localStorage so the runner can read it afterwards)
const journalKey = `repro.journal.${tag}`
const journal = (event: string, extra: Record<string, unknown> = {}) => {
  try {
    const list = JSON.parse(localStorage.getItem(journalKey) ?? "[]")
    list.push({ t: Date.now(), event, href: location.pathname + location.search, ...extra })
    localStorage.setItem(journalKey, JSON.stringify(list))
  } catch {}
}
// One id per loaded document, so every journal entry (and every SDK console message, see below)
// can be attributed to the document that produced it — even messages printed while it unloads.
const docId = Math.random().toString(36).slice(2, 8)
let pagehideFired = false
journal("script-start", { docId, variant, origin, autoFlushInterval, replayEnabled })
addEventListener("pagehide", (e) => {
  pagehideFired = true
  journal("pagehide(page listener, registered before SDK)", { docId, persisted: (e as PageTransitionEvent).persisted })
})
document.addEventListener("visibilitychange", () => journal(`visibilitychange:${document.visibilityState}`, { docId }))

// Mirror SDK console warnings/errors into the localStorage journal synchronously, tagged with the
// document id and whether pagehide already fired in this document. (Installed before the SDK, so
// the replay engine's own console wrapper later wraps this one; behaviour is unchanged.)
for (const level of ["warn", "error"] as const) {
  const orig = console[level]
  console[level] = (...args: unknown[]) => {
    const first = typeof args[0] === "string" ? args[0] : ""
    if (first.startsWith("[MapleClientSDK]") || first.startsWith("[maple]"))
      journal(`console.${level}`, { docId, afterPagehide: pagehideFired, visibility: document.visibilityState, text: `${first} ${args.slice(1).map((a) => String(a)).join(" ")}`.slice(0, 300) })
    orig.apply(console, args)
  }
}

// Optional user-land workaround (?shim=1): installed BEFORE the SDK. Any keepalive POST whose body
// would not fit Chromium's 64 KiB keepalive budget is sent as a normal (non-keepalive) request instead
// of being rejected outright. This does not help during unload, but it stops one oversized batch from
// disabling export for the rest of the page's life. (NOT part of the SDK.)
if (params.get("shim") === "1") {
  const origFetch = window.fetch.bind(window)
  const LIMIT = 60 * 1024
  window.fetch = (input: RequestInfo | URL, init?: RequestInit) => {
    if (init?.keepalive && typeof init.body === "string" && new TextEncoder().encode(init.body).byteLength > LIMIT) {
      journal("shim: keepalive downgraded", { bytes: new TextEncoder().encode(init.body).byteLength })
      return origFetch(input, { ...init, keepalive: false })
    }
    return origFetch(input, init)
  }
}

const common = {
  serviceName: "repro-web",
  environment: "repro",
  endpoint,
  ingestKey: "~", // placeholder, exactly like a same-origin proxy that injects the real key server-side
  replay: { enabled: replayEnabled },
}

let flushFn: (() => Promise<void>) | undefined
let layer: Layer.Layer<any, any, never>

if (variant === "flush") {
  const telemetry = MapleFlush.make({ ...common, ...(autoFlushInterval !== undefined ? { autoFlushInterval } : {}) })
  flushFn = telemetry.flush
  layer = telemetry.layer
} else if (variant === "layer") {
  layer = Maple.layer(common)
} else if (variant === "layer-flusher") {
  layer = Maple.layer(common).pipe(Layer.provide(Layer.succeed(FetchHttpClient.RequestInit, { keepalive: true })))
} else {
  throw new Error(`unknown variant ${variant}`)
}

const runtime = ManagedRuntime.make(layer)

if (variant === "layer-flusher") {
  const flushAll = Effect.gen(function* () {
    const flusher = yield* OtlpExporter.Flusher
    yield* flusher.flush
  })
  flushFn = () => runtime.runPromise(flushAll as Effect.Effect<void, never, never>)
  addEventListener("pagehide", () => void flushFn!())
  document.addEventListener("visibilitychange", () => {
    if (document.visibilityState === "hidden") void flushFn!()
  })
}

const counter = Metric.counter("repro_counter")

const randomText = (n: number) => {
  // random (incompressible-ish) text, so gzip'd replay chunks stay large
  const alphabet = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
  let s = ""
  const buf = new Uint8Array(n)
  crypto.getRandomValues(buf)
  for (let i = 0; i < n; i++) s += alphabet[buf[i]! % alphabet.length]
  return s
}

const T = {
  tag,
  docId,
  variant,
  endpoint,
  ready: false,
  /** Create `n` sibling spans named `${prefix}-${i}`, each with a `pad` attribute of `attrBytes` chars. */
  spans: async (n: number, attrBytes = 0, prefix = "span") => {
    const pad = "x".repeat(attrBytes)
    await runtime.runPromise(
      Effect.forEach(
        Array.from({ length: n }, (_, i) => i),
        (i) => Effect.void.pipe(Effect.withSpan(`${prefix}-${i}`, { attributes: attrBytes > 0 ? { pad } : {} })),
        { discard: true },
      ),
    )
    journal("spans-created", { n, attrBytes, prefix })
    return Date.now()
  },
  /** One span with an explicit name. */
  span: async (name: string, attrBytes = 0) => {
    await runtime.runPromise(Effect.void.pipe(Effect.withSpan(name, { attributes: attrBytes > 0 ? { pad: "x".repeat(attrBytes) } : {} })))
    journal("span-created", { name })
    return Date.now()
  },
  logs: async (n: number, bytes = 0, prefix = "log") => {
    const pad = "y".repeat(bytes)
    await runtime.runPromise(
      Effect.forEach(
        Array.from({ length: n }, (_, i) => i),
        (i) => Effect.logInfo(`${prefix}-${i} ${pad}`),
        { discard: true },
      ),
    )
    journal("logs-created", { n, bytes, prefix })
    return Date.now()
  },
  /** Effect log with several message arguments (Effect.logError(a, b, c, obj)). */
  logMulti: async (prefix = "multi") => {
    await runtime.runPromise(Effect.logError(`${prefix} Failed to connect`, new Error("ClientError: boom"), "Unknown Error", { code: 42, nested: { a: 1 } }))
    return Date.now()
  },
  metric: async () => {
    await runtime.runPromise(Metric.update(counter, 1))
    return Date.now()
  },
  flush: async () => {
    if (!flushFn) return "no-flush-fn"
    await flushFn()
    return "flushed"
  },
  /** Append `bytes` of random text to the DOM (drives rrweb incremental-snapshot size). */
  mutate: (bytes: number, chunks = 1) => {
    const host = document.getElementById("host")!
    for (let c = 0; c < chunks; c++) {
      const div = document.createElement("div")
      div.textContent = randomText(Math.floor(bytes / chunks))
      host.appendChild(div)
    }
    return Date.now()
  },
  journal: () => JSON.parse(localStorage.getItem(journalKey) ?? "[]"),
  clearJournal: () => localStorage.removeItem(journalKey),
}

;(globalThis as any).T = T

// Build the runtime eagerly so the tracer/logger are installed before the runner acts.
runtime.runPromise(Effect.void).then(() => {
  T.ready = true
  journal("runtime-ready")
  document.getElementById("status")!.textContent = `ready: variant=${variant} endpoint=${endpoint}`
})

document.getElementById("btn-spans")?.addEventListener("click", () => void T.spans(50, 500, "click"))
document.getElementById("btn-reload")?.addEventListener("click", () => location.reload())
@@@@@@@@@@ END FILE: repro/client/app.ts

@@@@@@@@@@ FILE: repro/client/probe.ts (2686 bytes)
/**
 * SDK-independent probe of Chromium's fetch(keepalive) limits.
 * Exposes `window.P`. Every request goes to the mock server's /probe/sink,
 * which records arrival + byte size and can delay its response (?delay=ms) so a
 * request stays "in flight" for as long as we want.
 */
const params = new URLSearchParams(location.search)
const tag = params.get("tag") ?? "probe"
const crossPort = params.get("crossPort") ?? "4174"

export interface SendOptions {
  readonly delay?: number
  readonly keepalive?: boolean
  readonly cross?: boolean
  readonly contentType?: string
  readonly id?: string
  /** Use a body made of a 2-byte UTF-8 character ("é") instead of ASCII, so JS string.length = bytes / 2. */
  readonly utf8TwoByte?: boolean
  /** Read the response body (await res.text()) before resolving. The SDK's keepalive transport does NOT do this. */
  readonly consume?: boolean
}

const send = async (bytes: number, opts: SendOptions = {}) => {
  const id = opts.id ?? `p${Math.random().toString(36).slice(2, 8)}`
  const base = opts.cross ? `http://127.0.0.1:${crossPort}` : location.origin
  const url = `${base}/probe/sink?tag=${encodeURIComponent(tag)}&id=${id}&delay=${opts.delay ?? 0}`
  const body = opts.utf8TwoByte ? "é".repeat(Math.floor(bytes / 2)) : "x".repeat(bytes)
  const t0 = performance.now()
  try {
    const res = await fetch(url, {
      method: "POST",
      keepalive: opts.keepalive ?? true,
      headers: opts.contentType ? { "content-type": opts.contentType } : undefined,
      body,
    })
    if (opts.consume) await res.text()
    return { id, bytes, jsStringLength: body.length, ok: res.ok, status: res.status, consumed: !!opts.consume, ms: Math.round(performance.now() - t0) }
  } catch (error) {
    return { id, bytes, jsStringLength: body.length, ok: false, error: String(error), ms: Math.round(performance.now() - t0) }
  }
}

const sleep = (ms: number) => new Promise((r) => setTimeout(r, ms))

/** Arm a pagehide handler that fires the given keepalive posts (no await possible during unload). */
const armUnloadBurst = (sizes: number[], opts: SendOptions = {}) => {
  addEventListener("pagehide", () => {
    sizes.forEach((bytes, i) => {
      void send(bytes, { ...opts, id: `${opts.id ?? "unload"}-${i}` }).then((r) => {
        try {
          const prev = JSON.parse(localStorage.getItem(`probe.unload.${tag}`) ?? "[]")
          prev.push(r)
          localStorage.setItem(`probe.unload.${tag}`, JSON.stringify(prev))
        } catch {}
      })
    })
  })
  return "armed"
}

;(globalThis as any).P = { tag, send, sleep, armUnloadBurst, ready: true }
document.getElementById("status")!.textContent = `probe ready (tag=${tag})`
@@@@@@@@@@ END FILE: repro/client/probe.ts

@@@@@@@@@@ FILE: repro/lib.ts (7349 bytes)
/**
 * Shared harness for run-scenarios.ts: browser launch, per-run capture of
 * browser console + Playwright network events + server-side ingest records,
 * and evidence writing.
 */
import { chromium, type Browser, type BrowserContext, type Page } from "playwright"
import { mkdirSync, writeFileSync, appendFileSync } from "node:fs"
import { join } from "node:path"
import { startServer, type IngestRecord, type Rule } from "./server.ts"

export const PAGE_PORT = Number(process.env.PAGE_PORT ?? 4173)
export const CROSS_PORT = Number(process.env.CROSS_PORT ?? 4174)
export const EVIDENCE = process.env.EVIDENCE_DIR ?? "./evidence-out"
/** "shell" = Playwright default (chromium-headless-shell); "new" = full Chromium in new headless mode (channel "chromium"). */
export const HEADLESS = process.env.HEADLESS ?? "shell"

mkdirSync(EVIDENCE, { recursive: true })
export const server = startServer({ pagePort: PAGE_PORT, crossPort: CROSS_PORT, logFile: join(EVIDENCE, "server-requests-all.jsonl") })

export const launch = (mode: string = HEADLESS): Promise<Browser> => chromium.launch(mode === "new" ? { channel: "chromium" } : {})

export const sleep = (ms: number) => new Promise((r) => setTimeout(r, ms))

let tagCounter = 0
export const newTag = (scenario: string) => `${scenario}-${Date.now().toString(36)}-${(tagCounter++).toString(36)}`

export interface ConsoleEntry {
  t: number
  type: string
  text: string
  page: string
}
export interface PwEntry {
  t: number
  ev: "request" | "finished" | "failed"
  method: string
  url: string
  postBytes?: number
  failure?: string | null
  status?: number
}

const shortUrl = (u: string) => u.replace(`http://localhost:${PAGE_PORT}`, "").replace(`http://127.0.0.1:${CROSS_PORT}`, "[cross]")

export class Run {
  readonly t0 = Date.now()
  readonly console: ConsoleEntry[] = []
  readonly pw: PwEntry[] = []
  readonly marks: { t: number; name: string; [k: string]: unknown }[] = []
  constructor(
    readonly scenario: string,
    readonly label: string,
    readonly index: number,
    readonly tag: string,
    readonly config: Record<string, unknown> = {},
  ) {}

  rel(t: number = Date.now()) {
    return t - this.t0
  }

  mark(name: string, extra: Record<string, unknown> = {}) {
    this.marks.push({ t: this.rel(), name, ...extra })
  }

  attach(page: Page) {
    page.on("console", (m) => this.console.push({ t: this.rel(), type: m.type(), text: m.text(), page: shortUrl(page.url()) }))
    page.on("pageerror", (e) => this.console.push({ t: this.rel(), type: "pageerror", text: String(e), page: shortUrl(page.url()) }))
    page.on("request", (r) => {
      if (!/\/otlp\/|\/probe\//.test(r.url())) return
      let postBytes: number | undefined
      try {
        postBytes = r.postDataBuffer()?.length
      } catch {}
      this.pw.push({ t: this.rel(), ev: "request", method: r.method(), url: shortUrl(r.url()), postBytes })
    })
    page.on("requestfinished", async (r) => {
      if (!/\/otlp\/|\/probe\//.test(r.url())) return
      const t = this.rel()
      let status: number | undefined
      try {
        status = (await r.response())?.status()
      } catch {}
      this.pw.push({ t, ev: "finished", method: r.method(), url: shortUrl(r.url()), status })
    })
    page.on("requestfailed", (r) => {
      if (!/\/otlp\/|\/probe\//.test(r.url())) return
      this.pw.push({ t: this.rel(), ev: "failed", method: r.method(), url: shortUrl(r.url()), failure: r.failure()?.errorText ?? null })
    })
  }

  records(): IngestRecord[] {
    return server.records.filter((r) => r.tag === this.tag)
  }

  /** Every span name that reached the ingest for this run (POSTs answered 200 only). */
  arrivedSpans(): Map<string, number> {
    const out = new Map<string, number>()
    for (const r of this.records()) {
      const s = r.summary as any
      if (r.method !== "POST" || s?.kind !== "traces" || r.status !== 200) continue
      for (const n of s.spanNames) out.set(n, this.rel(r.tRecv))
    }
    return out
  }

  arrivedLogs(): Map<string, number> {
    const out = new Map<string, number>()
    for (const r of this.records()) {
      const s = r.summary as any
      if (r.method !== "POST" || s?.kind !== "logs" || r.status !== 200) continue
      for (const b of s.logBodies) out.set(String(b).split(" ")[0]!, this.rel(r.tRecv))
    }
    return out
  }

  consoleMatching(re: RegExp) {
    return this.console.filter((c) => re.test(c.text))
  }

  async waitForRecord(pred: (r: IngestRecord) => boolean, timeoutMs = 15_000): Promise<IngestRecord | undefined> {
    const end = Date.now() + timeoutMs
    while (Date.now() < end) {
      const hit = this.records().find(pred)
      if (hit) return hit
      await sleep(25)
    }
    return undefined
  }

  save(result: Record<string, unknown>, extra: Record<string, unknown> = {}) {
    const dir = join(EVIDENCE, this.scenario)
    mkdirSync(dir, { recursive: true })
    const file = join(dir, `${this.label}-run${this.index}.json`)
    const records = this.records().map((r) => ({
      ...r,
      tRecv: this.rel(r.tRecv),
      tBody: r.tBody !== undefined ? this.rel(r.tBody) : undefined,
      tResp: r.tResp !== undefined ? this.rel(r.tResp) : undefined,
    }))
    writeFileSync(
      file,
      JSON.stringify(
        {
          scenario: this.scenario,
          label: this.label,
          run: this.index,
          tag: this.tag,
          note: "all t* values are milliseconds relative to the start of this run",
          config: this.config,
          result,
          marks: this.marks,
          browserConsole: this.console,
          playwrightNetwork: this.pw,
          serverRecords: records,
          ...extra,
        },
        null,
        2,
      ),
    )
    return file
  }
}

export const appUrl = (tag: string, q: Record<string, string | number | boolean>) => {
  const p = new URLSearchParams({ tag, crossPort: String(CROSS_PORT) })
  for (const [k, v] of Object.entries(q)) p.set(k, String(v))
  return `http://localhost:${PAGE_PORT}/?${p}`
}

export const openApp = async (ctx: BrowserContext, run: Run, q: Record<string, string | number | boolean>) => {
  const page = await ctx.newPage()
  run.attach(page)
  run.mark("goto")
  await page.goto(appUrl(run.tag, q))
  await page.waitForFunction(() => (globalThis as any).T?.ready === true, undefined, { timeout: 15_000 })
  run.mark("ready")
  return page
}

export const waitReady = async (page: Page) => page.waitForFunction(() => (globalThis as any).T?.ready === true, undefined, { timeout: 15_000 })

/** Add failure/latency rules scoped to one run's tag (so runs can execute in parallel). */
export const addRulesFor = (run: Run, rules: Omit<Rule, "tag">[]) => {
  server.addRules(rules.map((r) => ({ ...r, tag: run.tag })))
  run.config.rules = [...((run.config.rules as unknown[]) ?? []), ...rules]
}
export const clearRulesFor = (run: Run) => server.clearRulesForTag(run.tag)

export interface SummaryRow {
  scenario: string
  label: string
  run: number
  verdict: string
  details: string
  evidence: string
}
export const summary: SummaryRow[] = []
export const addSummary = (row: SummaryRow) => {
  summary.push(row)
  appendFileSync(join(EVIDENCE, "summary.jsonl"), `${JSON.stringify(row)}\n`)
  console.log(`[${row.scenario}/${row.label}#${row.run}] ${row.verdict} — ${row.details}`)
}
@@@@@@@@@@ END FILE: repro/lib.ts

@@@@@@@@@@ FILE: repro/public/index.html (629 bytes)
<!doctype html>
<html>
  <head>
    <meta charset="utf-8" />
    <title>maple effect-sdk keepalive repro</title>
  </head>
  <body>
    <h1>maple effect-sdk keepalive repro</h1>
    <p id="status">loading…</p>
    <button id="btn-spans">create 50 spans × 500 B</button>
    <button id="btn-reload">reload</button>
    <div id="host"></div>
    <script type="module">
      // ?sdk=0.10.0 loads the optional build against the latest SDK (see build.ts); default is 0.8.2
      const sdk = new URLSearchParams(location.search).get("sdk")
      import(sdk ? `/dist-${sdk}/app.js` : "/dist/app.js")
    </script>
  </body>
</html>
@@@@@@@@@@ END FILE: repro/public/index.html

@@@@@@@@@@ FILE: repro/public/probe.html (293 bytes)
<!doctype html>
<html>
  <head>
    <meta charset="utf-8" />
    <title>keepalive probe (no SDK)</title>
  </head>
  <body>
    <h1>fetch(keepalive) probe — no SDK on this page</h1>
    <p id="status">loading…</p>
    <script type="module" src="/dist/probe.js"></script>
  </body>
</html>
@@@@@@@@@@ END FILE: repro/public/probe.html

@@@@@@@@@@ FILE: repro/run-scenarios.ts (8382 bytes)
/**
 * Playwright runner for all scenarios.
 *
 *   bun run run-scenarios.ts                 # everything (≈ 25 min)
 *   bun run run-scenarios.ts B C D           # only some groups
 *   REPEAT=1 bun run run-scenarios.ts E      # override repeat count for every group
 *   HEADLESS=new bun run run-scenarios.ts G  # full Chromium "new headless" instead of headless-shell
 *   EVIDENCE_DIR=../evidence bun run run-scenarios.ts
 *
 * Groups: B (probe + SDK concurrency), A0, A, C, D, Dr, E, F, G, H, I, K, L, W
 */
import { writeFileSync } from "node:fs"
import { join } from "node:path"
import { chromium } from "playwright"
import { EVIDENCE, HEADLESS, launch, server, summary } from "./lib.ts"
import { scenarioBProbe } from "./scenarios/b-probe.ts"
import * as S from "./scenarios/sdk.ts"

const args = process.argv.slice(2)
const groups = args.length > 0 ? new Set(args) : new Set(["B", "A0", "A", "C", "D", "Dr", "E", "F", "G", "H", "I", "K", "L", "W"])
// SDK=0.10.0 runs the same scenarios against the optional latest-SDK build (see build.ts)
if (process.env.SDK) S.setExtraQuery({ sdk: process.env.SDK })
const REPEAT = process.env.REPEAT ? Number(process.env.REPEAT) : undefined
const rep = (n: number) => REPEAT ?? n

const startedAt = new Date()
const browser = await launch()
const env = {
  sdkUnderTest: process.env.SDK ? `@maple-dev/effect-sdk@${process.env.SDK} (optional variant build)` : "@maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112",
  startedAt: startedAt.toISOString(),
  headlessMode: HEADLESS === "new" ? "new headless (full Chromium, channel 'chromium')" : "chromium-headless-shell (Playwright default)",
  chromiumVersion: browser.version(),
  playwrightVersion: (await import("playwright/package.json", { with: { type: "json" } })).default.version,
  bunVersion: Bun.version,
  platform: `${process.platform} ${process.arch}`,
  userAgent: await (async () => {
    const p = await browser.newPage()
    const ua = await p.evaluate(() => navigator.userAgent)
    await p.close()
    return ua
  })(),
  groups: [...groups],
}
console.log("environment:", env)
writeFileSync(join(EVIDENCE, `environment-${HEADLESS}${process.env.SDK ? `-sdk${process.env.SDK}` : ""}-${args.join("_") || "all"}.json`), JSON.stringify(env, null, 2))

const times = async (n: number, fn: (i: number) => Promise<unknown>) => {
  for (let i = 0; i < n; i++) await fn(i)
}

if (groups.has("B")) {
  await times(rep(2), (i) => scenarioBProbe(browser, i))
  await times(rep(3), (i) => S.Bsdk(browser, i, "manual-flush", "flush"))
  await times(rep(3), (i) => S.Bsdk(browser, i, "reload", "reload"))
}

if (groups.has("A0")) {
  for (const n of [100, 150, 160, 163, 164, 165, 170, 200]) await times(rep(1), (i) => S.A0(browser, i, n))
}

if (groups.has("A")) {
  // long-running (≈ 2.5 min each); every run gets its own browser process and they run in parallel
  const variants: [string, Record<string, string | number>, Parameters<typeof S.A>[4]][] = [
    ["burst200x500B-interval5000(default)", {}, { burstSpans: 200, burstAttrBytes: 500, observeMs: 130_000 }],
    ["burst200x500B-interval1000", { interval: 1000 }, { burstSpans: 200, burstAttrBytes: 500, observeMs: 130_000 }],
    ["steady30spansPerSec-10s-interval5000(default)", {}, { burstSpans: 0, burstAttrBytes: 0, steadyPerSec: 30, steadySecs: 10, observeMs: 75_000 }],
    ["steady40spansPerSec-10s-interval5000(default)", {}, { burstSpans: 0, burstAttrBytes: 0, steadyPerSec: 40, steadySecs: 10, observeMs: 75_000 }],
  ]
  const jobs: Promise<unknown>[] = []
  for (const [label, q, opts] of variants)
    for (let i = 0; i < rep(2); i++)
      jobs.push(
        (async () => {
          const b = await launch()
          try {
            await S.A(b, i, label, q, opts)
          } finally {
            await b.close()
          }
        })(),
      )
  await Promise.all(jobs)
}

if (groups.has("C")) {
  await times(rep(3), (i) => S.C(browser, i, "inject503-then-reload", "reload", true))
  await times(rep(3), (i) => S.C(browser, i, "inject503-then-navigate", "navigate", true))
  await times(rep(3), (i) => S.C(browser, i, "inject503-then-close", "close", true))
  await times(rep(3), (i) => S.C(browser, i, "control-no-failure-reload", "reload", false))
}

if (groups.has("D")) {
  await times(rep(3), (i) => S.D(browser, i, "first-POST-4000ms-interval1000", 4000, 1000))
  await times(rep(3), (i) => S.D(browser, i, "first-POST-4000ms-interval5000", 4000, 5000))
  await times(rep(3), (i) => S.D(browser, i, "first-POST-7000ms-interval1000-tail-5s-before-reload", 7000, 1000, 5000))
  await times(rep(3), (i) => S.D(browser, i, "control-no-latency-interval1000", 0, 1000))
  await times(rep(2), (i) => S.Dhang(browser, i, "first-POST-held-15s", 15000))
}

if (groups.has("Dr")) {
  await times(rep(10), (i) => S.Dr(browser, i, "latency150-interval1000", 150, 1000))
  await times(rep(10), (i) => S.Dr(browser, i, "latency600-interval1000", 600, 1000))
  await times(rep(10), (i) => S.Dr(browser, i, "latency600-interval5000", 600, 5000))
  await times(rep(10), (i) => S.Dr(browser, i, "control-latency0-interval1000", 0, 1000))
}

if (groups.has("E")) {
  await times(rep(3), (i) => S.E(browser, i, "E1-spans45-plus-60kB-DOM-mutation", {}, { spans: 45, spanAttr: 0, mutateChars: 60000, consoleLines: 0 }))
  await times(rep(3), (i) => S.E(browser, i, "E2-spans150(~60kB)-plus-30-console-lines", {}, { spans: 150, spanAttr: 0, mutateChars: 0, consoleLines: 30 }))
  await times(rep(3), (i) => S.E(browser, i, "E3-replay-disabled-spans150(~60kB)", { replay: 0 }, { spans: 150, spanAttr: 0, mutateChars: 0, consoleLines: 0 }))
  await times(rep(3), (i) => S.E(browser, i, "E4-no-spans-small-DOM-mutation", {}, { spans: 0, spanAttr: 0, mutateChars: 2000, consoleLines: 3 }))
  await times(rep(3), (i) => S.E(browser, i, "E5-Maple.layer-small-DOM-mutation", { variant: "layer" }, { spans: 0, spanAttr: 0, mutateChars: 2000, consoleLines: 3 }))
}

if (groups.has("F")) {
  for (const v of ["flush", "layer"]) {
    await times(rep(2), (i) => S.F(browser, i, `${v}-periodic-events-POST-in-flight`, v, "events"))
    await times(rep(2), (i) => S.F(browser, i, `${v}-active-meta-POST-in-flight`, v, "meta"))
    await times(rep(2), (i) => S.Fr(browser, i, `${v}-8-random-reloads-latency300`, v, 300))
  }
}

if (groups.has("G")) {
  await times(rep(2), (i) => S.G(browser, i, "steady-state-no-navigation", false))
  await times(rep(2), (i) => S.G(browser, i, "with-final-reload", true))
}

if (groups.has("H")) {
  for (const v of ["layer", "layer-flusher", "flush"]) await times(rep(3), (i) => S.H(browser, i, v, v))
}

if (groups.has("I")) {
  await times(rep(2), (i) => S.Icross(browser, i, "same-origin", "same"))
  await times(rep(2), (i) => S.Icross(browser, i, "cross-origin", "cross"))
}

if (groups.has("K")) {
  await times(rep(3), (i) => S.K(browser, i, "replay-on(default)", {}))
  await times(rep(3), (i) => S.K(browser, i, "replay-off", { replay: 0 }))
}

if (groups.has("L")) await times(rep(1), (i) => S.L(browser, i))

if (groups.has("W")) {
  await times(rep(2), (i) => S.W(browser, i, "without-shim", false))
  await times(rep(2), (i) => S.W(browser, i, "with-keepalive-downgrade-shim", true))
}

await browser.close()
server.stop()

// ---- summary
const grouped = new Map<string, { verdicts: Map<string, number>; details: string[] }>()
for (const row of summary) {
  const key = `${row.scenario} / ${row.label}`
  const g = grouped.get(key) ?? { verdicts: new Map(), details: [] }
  g.verdicts.set(row.verdict, (g.verdicts.get(row.verdict) ?? 0) + 1)
  g.details.push(`  run ${row.run}: ${row.verdict} — ${row.details.split("\n")[0]}\n      evidence: ${row.evidence}`)
  grouped.set(key, g)
}
let out = `Scenario summary (${env.headlessMode}, Chromium ${env.chromiumVersion}, Playwright ${env.playwrightVersion}, Bun ${env.bunVersion})\nstarted ${env.startedAt}, finished ${new Date().toISOString()}\n\n`
for (const [key, g] of grouped) {
  const total = [...g.verdicts.values()].reduce((a, b) => a + b, 0)
  out += `${key}  [${[...g.verdicts.entries()].map(([v, n]) => `${v}: ${n}/${total}`).join("; ")}]\n${g.details.join("\n")}\n\n`
}
const file = join(EVIDENCE, `summary-${HEADLESS}${process.env.SDK ? `-sdk${process.env.SDK}` : ""}-${startedAt.toISOString().replace(/[:.]/g, "-")}.txt`)
writeFileSync(file, out)
console.log(`\n${out}\nsummary written to ${file}`)
process.exit(0)
@@@@@@@@@@ END FILE: repro/run-scenarios.ts

@@@@@@@@@@ FILE: repro/scenarios/b-probe.ts (9438 bytes)
/**
 * Scenario B (part 1): SDK-independent measurement of Chromium's fetch(keepalive)
 * body limits, using /probe.html (no Maple SDK on the page).
 *
 * Every sub-test starts from a freshly loaded document so that budget held by a
 * previous sub-test cannot leak into the next one.
 */
import type { Browser, Page } from "playwright"
import { PAGE_PORT, Run, addSummary, newTag, sleep } from "../lib.ts"

const probeUrl = (tag: string) => `http://localhost:${PAGE_PORT}/probe.html?tag=${tag}`

const arrivedIds = (run: Run, prefix: string) =>
  run
    .records()
    .filter((r) => r.method === "POST" && r.query.includes(`id=${prefix}`))
    .map((r) => ({ id: new URLSearchParams(r.query).get("id"), bytes: r.bytes, t: run.rel(r.tRecv) }))

const ev = <T>(page: Page, fn: string): Promise<T> => page.evaluate(`(async () => { const P = globalThis.P; ${fn} })()`) as Promise<T>

export const scenarioBProbe = async (browser: Browser, index: number) => {
  const tag = newTag("B-probe")
  const run = new Run("B-keepalive-probe", "probe", index, tag, { page: "/probe.html (no SDK)" })
  const ctx = await browser.newContext()
  const page = await ctx.newPage()
  run.attach(page)
  const fresh = async (p: Page = page) => {
    await p.goto(probeUrl(tag))
    await p.waitForFunction(() => (globalThis as any).P?.ready)
  }
  const result: Record<string, unknown> = {}
  const ok = (r: any) => (r.ok ? "ok" : "REJECTED")

  // B1: single request, nothing else in flight (fresh document per size)
  const b1: unknown[] = []
  for (const size of [1024, 49152, 65535, 65536, 65537, 70000, 131072]) {
    await fresh()
    b1.push(await ev(page, `return P.send(${size}, { id: "B1-${size}", consume: true })`))
  }
  result.B1_single = b1
  await fresh()
  result.B1b_single_json_65536_then_65537 = await ev(page, `return [await P.send(65536, { id: "B1b-65536", contentType: "application/json", consume: true }), await P.send(65537, { id: "B1b-65537", contentType: "application/json", consume: true })]`)

  // B2: concurrent requests whose SUM exceeds 64 KiB although each is small; the sink
  // holds each response 2 s so they are all in flight together.
  await fresh()
  const b2a = await ev<any[]>(page, `return Promise.all([P.send(40000, { delay: 2000, id: "B2a-0" }), P.send(40000, { delay: 2000, id: "B2a-1" })])`)
  await fresh()
  const b2b = await ev<any[]>(page, `return Promise.all([P.send(30000, { delay: 2000, id: "B2b-0" }), P.send(30000, { delay: 2000, id: "B2b-1" })])`)
  await fresh()
  const b2c = await ev<any[]>(page, `return Promise.all([P.send(30000, { delay: 2000, id: "B2c-0" }), P.send(30000, { delay: 2000, id: "B2c-1" }), P.send(10000, { delay: 2000, id: "B2c-2" })])`)
  await fresh()
  const b2d = await ev<any[]>(page, `return Promise.all([P.send(20000, { delay: 2000, id: "B2d-0" }), P.send(20000, { delay: 2000, id: "B2d-1" }), P.send(20000, { delay: 2000, id: "B2d-2" }), P.send(20000, { delay: 2000, id: "B2d-3" })])`)
  result.B2_concurrent = { two_x_40000: b2a, two_x_30000: b2b, three_30000_30000_10000: b2c, four_x_20000: b2d }

  // B3: release timing. Hold 60 kB for 2 s; 10 kB at +500 ms; 10 kB after the hold completes.
  await fresh()
  result.B3_release_timing = await ev(page, `
    const hold = P.send(60000, { delay: 2000, id: "B3-hold60000", consume: true })
    await P.sleep(500)
    const during = await P.send(10000, { id: "B3-during", consume: true })
    const held = await hold
    const after = await P.send(10000, { id: "B3-after", consume: true })
    return { held, during_plus500ms: during, after_hold_completed: after }`)

  // B3b: budget is held until the response BODY finishes loading, not until fetch() resolves.
  // 8 sequential 40000 B keepalive POSTs, each awaited; variants: body consumed or not, gap between.
  const seq = async (opts: string, gapMs: number) => {
    await fresh()
    return ev<string>(page, `const out = []; for (let i = 0; i < 8; i++) { const r = await P.send(40000, ${opts}); out.push(r.ok ? "ok" : "X"); if (${gapMs}) await P.sleep(${gapMs}) } return out.join(" ")`)
  }
  result.B3b_sequential_8x40000 = {
    "response body NOT read, no gap (what the SDK transport does)": await seq(`{}`, 0),
    "response body read (await res.text()), no gap": await seq(`{ consume: true }`, 0),
    "response body NOT read, 50 ms gap": await seq(`{}`, 50),
  }

  // B4: cross-origin (CORS preflight, content-type: application/json)
  await fresh()
  const b4single = await ev(page, `return P.send(1000, { cross: true, contentType: "application/json", id: "B4-single", consume: true })`)
  await fresh()
  const b4over = await ev(page, `return P.send(65537, { cross: true, contentType: "application/json", id: "B4-over" })`)
  await fresh()
  const b4conc = await ev(page, `return Promise.all([P.send(40000, { cross: true, contentType: "application/json", delay: 2000, id: "B4c-0" }), P.send(40000, { cross: true, contentType: "application/json", delay: 2000, id: "B4c-1" })])`)
  result.B4_cross_origin = { single_1000_json: b4single, single_65537_json: b4over, two_x_40000_json_concurrent: b4conc }

  // B5: control — without keepalive there is no size limit
  await fresh()
  result.B5_no_keepalive_200000 = await ev(page, `return P.send(200000, { keepalive: false, id: "B5", consume: true })`)

  // B6: the limit is in UTF-8 BYTES, not JS string length
  const b6: Record<string, unknown> = {}
  for (const bytes of [60000, 65536, 80000, 98304]) {
    await fresh()
    b6[`utf8_bytes_${bytes}_strlen_${bytes / 2}`] = await ev(page, `return P.send(${bytes}, { utf8TwoByte: true, id: "B6-${bytes}", consume: true })`)
  }
  result.B6_utf8 = b6

  // B7: scope of the budget — other tab (same context) / other context while tab 1 holds 60 kB
  await fresh()
  const page2 = await ctx.newPage()
  run.attach(page2)
  await fresh(page2)
  const otherCtx = await browser.newContext()
  const page3 = await otherCtx.newPage()
  run.attach(page3)
  await fresh(page3)
  const hold = ev(page, `return P.send(60000, { delay: 3000, id: "B7-tab1-hold60000", consume: true })`)
  await sleep(300)
  const sameTabSecond = await ev(page, `return P.send(10000, { id: "B7-tab1-second-10000", consume: true })`)
  const tab2 = await ev(page2, `return P.send(60000, { id: "B7-tab2-60000", consume: true })`)
  const tab3 = await ev(page3, `return P.send(60000, { id: "B7-otherContext-60000", consume: true })`)
  result.B7_scope = { tab1_hold_60000: await hold, tab1_second_10000_while_holding: sameTabSecond, tab2_same_context_60000_while_tab1_holds: tab2, tab3_other_context_60000_while_tab1_holds: tab3 }
  await page2.close()
  await otherCtx.close()

  // B8: keepalive posts fired from a pagehide handler during a reload (fresh doc each)
  const unloadCase = async (name: string, sizes: number[]) => {
    await fresh()
    await page.evaluate(([sizes, name]) => (globalThis as any).P.armUnloadBurst(sizes, { id: name }), [sizes, name] as const)
    run.mark(`B8 reload (${name})`)
    await page.reload()
    await page.waitForFunction(() => (globalThis as any).P?.ready)
    await sleep(1500)
    const arrived = arrivedIds(run, `${name}-`)
    return { fired: sizes, firedTotal: sizes.reduce((a, b) => a + b, 0), arrivedAtServer: arrived, arrivedCount: arrived.length }
  }
  result.B8_unload_burst = {
    two_x_20000: await unloadCase("B8a", [20000, 20000]),
    two_x_40000: await unloadCase("B8b", [40000, 40000]),
    three_x_20000: await unloadCase("B8c", [20000, 20000, 20000]),
    four_x_20000: await unloadCase("B8d", [20000, 20000, 20000, 20000]),
    one_x_70000: await unloadCase("B8e", [70000]),
  }

  // B9: an in-flight keepalive from the PREVIOUS document does not count against the new document
  await fresh()
  await page.evaluate(() => void (globalThis as any).P.send(60000, { delay: 4000, id: "B9-prevdoc-hold60000" }))
  await sleep(200)
  run.mark("B9 reload while 60000 B keepalive in flight")
  await page.reload()
  await page.waitForFunction(() => (globalThis as any).P?.ready)
  result.B9_new_document_60000 = await ev(page, `return P.send(60000, { id: "B9-newdoc-60000", consume: true })`)
  await sleep(4500)
  result.B9_arrived = arrivedIds(run, "B9-")

  // B10: an in-flight keepalive in the SAME document blocks the pagehide keepalive
  await fresh()
  await page.evaluate(() => {
    const P = (globalThis as any).P
    void P.send(50000, { delay: 4000, id: "B10-periodic-hold50000" })
    P.armUnloadBurst([20000], { id: "B10-unload" })
  })
  await sleep(200)
  run.mark("B10 reload while 50000 B keepalive in flight, pagehide fires 20000 B")
  await page.reload()
  await page.waitForFunction(() => (globalThis as any).P?.ready)
  await sleep(4500)
  result.B10_arrived = arrivedIds(run, "B10-")

  await ctx.close()
  const file = run.save(result, { browserVersion: browser.version() })

  const single = (result.B1_single as any[]).map((r) => `${r.bytes}:${ok(r)}`).join(" ")
  addSummary({
    scenario: "B-probe",
    label: "probe",
    run: index,
    verdict: "MEASURED",
    details: `single: ${single} | 2x40000 concurrent: ${b2a.map(ok).join("/")} | 4x20000 concurrent: ${b2d.map(ok).join("/")} | unload 2x40000 arrived ${(result.B8_unload_burst as any).two_x_40000.arrivedCount}/2, 4x20000 arrived ${(result.B8_unload_burst as any).four_x_20000.arrivedCount}/4 | B10 unload-while-inflight arrived ${(result.B10_arrived as any[]).map((a) => a.id).join(",")}`,
    evidence: file,
  })
  return result
}
@@@@@@@@@@ END FILE: repro/scenarios/b-probe.ts

@@@@@@@@@@ FILE: repro/scenarios/sdk.ts (34500 bytes)
/**
 * SDK-level scenarios (A, A0, Bsdk, C, D, Dr, E, F, G, H, I). Each scenario
 * function performs ONE run in its own browser context and returns a verdict.
 * See README.md for what each scenario tests.
 */
import type { Browser, Page } from "playwright"
import { Run, addRulesFor, addSummary, clearRulesFor, newTag, openApp, sleep, waitReady, PAGE_PORT } from "../lib.ts"

type Q = Record<string, string | number | boolean>

const FAIL_RE = /\[MapleClientSDK\] (\w+) flush failed; cooldown 60s/
const SKIP_RE = /\[MapleClientSDK\] (\w+) flush skipped \(cooldown/
const REPLAY_RE = /\[maple\] session replay (.+?) failed/

const reload = async (page: Page, run: Run) => {
  run.mark("reload-start")
  await page.reload()
  run.mark("reload-load-event")
  await waitReady(page)
  run.mark("reload-ready")
}

const consoleCounts = (run: Run) => ({
  flushFailed: run.consoleMatching(FAIL_RE).map((c) => ({ t: c.t, text: c.text })),
  flushSkipped: run.consoleMatching(SKIP_RE).map((c) => ({ t: c.t, text: c.text })),
  replayPostFailed: run.consoleMatching(REPLAY_RE).map((c) => ({ t: c.t, text: c.text })),
})

/** Extra query parameters added to every app URL (e.g. { sdk: "0.10.0" } for the optional latest-SDK build). */
let extraQuery: Q = {}
export const setExtraQuery = (q: Q) => {
  extraQuery = q
}

/** Read the page's localStorage journal (lifecycle events + SDK console messages tagged per document). */
const readJournal = async (ctx: import("playwright").BrowserContext, run: Run) => {
  try {
    const p = await ctx.newPage()
    await p.goto(`http://localhost:${PAGE_PORT}/probe.html?tag=${run.tag}-journal`)
    const j = await p.evaluate((key) => JSON.parse(localStorage.getItem(key) ?? "[]"), `repro.journal.${run.tag}`)
    await p.close()
    return (j as any[]).map((e) => ({ ...e, t: run.rel(e.t) }))
  } catch (error) {
    return [{ error: String(error) }]
  }
}

const withRun = async (
  browser: Browser,
  scenario: string,
  label: string,
  index: number,
  q: Q,
  body: (page: Page, run: Run) => Promise<{ verdict: string; details: string; result: Record<string, unknown> }>,
) => {
  q = { ...q, ...extraQuery }
  const run = new Run(scenario, label, index, newTag(scenario), { query: q, browserVersion: browser.version() })
  const ctx = await browser.newContext()
  try {
    const page = await openApp(ctx, run, q)
    const out = await body(page, run)
    const journal = await readJournal(ctx, run)
    const file = run.save({ ...out.result, console: consoleCounts(run) }, { pageJournal: journal })
    addSummary({ scenario, label, run: index, verdict: out.verdict, details: out.details, evidence: file })
    return out
  } catch (error) {
    const file = run.save({ harnessError: String(error) })
    addSummary({ scenario, label, run: index, verdict: "HARNESS-ERROR", details: String(error), evidence: file })
    return { verdict: "HARNESS-ERROR", details: String(error), result: {} }
  } finally {
    clearRulesFor(run)
    await ctx.close()
  }
}

/** Wait until a traces POST containing `spanName` has been received (headers+body) by the server. */
const waitSpanReceived = (run: Run, spanName: string, timeoutMs = 15_000) =>
  run.waitForRecord((r) => (r.summary as any)?.kind === "traces" && (r.summary as any).spanNames.includes(spanName) && r.tBody !== undefined, timeoutMs)

// ---------------------------------------------------------------------------
// A0: threshold — how many plain spans fit in one flush before the keepalive POST is rejected?
export const A0 = (browser: Browser, index: number, spans: number) =>
  withRun(browser, "A0-threshold", `n${spans}`, index, { variant: "flush", interval: "false" }, async (page, run) => {
    await page.evaluate((n) => (globalThis as any).T.spans(n, 0, "A0"), spans)
    run.mark("spans-created")
    await page.evaluate(() => (globalThis as any).T.flush())
    run.mark("flush-returned")
    await sleep(500)
    const arrived = [...run.arrivedSpans().keys()].filter((n) => n.startsWith("A0-")).length
    const traceReqs = run.pw.filter((p) => p.ev === "request" && p.url.includes("/v1/traces"))
    const failed = run.consoleMatching(FAIL_RE)
    return {
      verdict: arrived === spans ? "DELIVERED" : "REJECTED",
      details: `${spans} spans in one flush: arrived ${arrived}; traces POST body bytes (Playwright) ${traceReqs.map((r) => r.postBytes).join(",") || "n/a"}; console: ${failed.map((f) => f.text).join(" | ") || "none"}`,
      result: { spans, arrived, traceRequestBodyBytes: traceReqs.map((r) => r.postBytes) },
    }
  })

// ---------------------------------------------------------------------------
// A: an oversized batch poisons trace export for the lifetime of the page.
export const A = (browser: Browser, index: number, label: string, q: Q, opts: { burstSpans: number; burstAttrBytes: number; steadyPerSec?: number; steadySecs?: number; observeMs: number }) =>
  withRun(browser, "A-oversized-batch", label, index, { variant: "flush", ...q }, async (page, run) => {
    await sleep(1000)
    const created: string[] = []
    if (opts.burstSpans > 0) {
      await page.evaluate(([n, b]) => (globalThis as any).T.spans(n, b, "A-burst"), [opts.burstSpans, opts.burstAttrBytes] as const)
      for (let i = 0; i < opts.burstSpans; i++) created.push(`A-burst-${i}`)
      run.mark("burst-created", { spans: opts.burstSpans, attrBytes: opts.burstAttrBytes })
    }
    if (opts.steadyPerSec) {
      // steady stream, created in 100 ms slices
      const slices = (opts.steadySecs ?? 10) * 10
      const perSlice = opts.steadyPerSec / 10
      for (let s = 0; s < slices; s++) {
        await page.evaluate(([n, p]) => (globalThis as any).T.spans(n, 0, p), [perSlice, `A-steady-${s}`] as const)
        for (let i = 0; i < perSlice; i++) created.push(`A-steady-${s}-${i}`)
        await sleep(100)
      }
      run.mark("steady-done")
    }
    const start = Date.now()
    let k = 0
    while (Date.now() - start < opts.observeMs) {
      await page.evaluate((name) => (globalThis as any).T.span(name), `A-tick-${k}`)
      created.push(`A-tick-${k}`)
      if (k % 5 === 0) await page.evaluate((p) => (globalThis as any).T.logs(1, 0, p), `A-log-${k}`)
      k++
      await sleep(2000)
    }
    // allow the last regular flush to happen before judging
    await sleep(Number(q.interval ?? 5000) + 2000)
    const arrived = run.arrivedSpans()
    const arrivedCreated = created.filter((n) => arrived.has(n))
    const ticks = created.filter((n) => n.startsWith("A-tick-"))
    const ticksArrived = ticks.filter((n) => arrived.has(n))
    const logsArrived = [...run.arrivedLogs().keys()].filter((n) => n.startsWith("A-log-")).length
    const c = consoleCounts(run)
    const tracePosts = run.records().filter((r) => r.method === "POST" && r.path.endsWith("/v1/traces"))
    return {
      verdict: arrivedCreated.length === 0 && c.flushFailed.length > 0 ? "REPRODUCED (no span ever delivered)" : arrivedCreated.length < created.length ? "PARTIAL LOSS" : "NOT REPRODUCED",
      details: `created ${created.length} spans (${ticks.length} small 'tick' spans after the burst); delivered ${arrivedCreated.length} (ticks ${ticksArrived.length}/${ticks.length}); traces POSTs reaching server: ${tracePosts.length}; logs delivered ${logsArrived}; console 'flush failed' at t=${c.flushFailed.map((f) => f.t).join(",")}ms`,
      result: {
        createdCount: created.length,
        deliveredCount: arrivedCreated.length,
        ticksCreated: ticks.length,
        ticksDelivered: ticksArrived.length,
        logsDelivered: logsArrived,
        tracePostsAtServer: tracePosts.length,
        flushFailedAt: c.flushFailed.map((f) => f.t),
      },
    }
  })

// ---------------------------------------------------------------------------
// Bsdk: traces + logs + metrics are POSTed concurrently (Promise.all) as separate keepalive
// requests; each fits under 64 KiB but their sum does not.
export const Bsdk = (browser: Browser, index: number, label: string, trigger: "flush" | "reload") =>
  withRun(browser, "B-sdk-concurrent-signals", label, index, { variant: "flush", interval: trigger === "flush" ? "false" : 60000 }, async (page, run) => {
    await page.evaluate(() => (globalThis as any).T.spans(40, 500, "Bsdk-span"))
    await page.evaluate(() => (globalThis as any).T.logs(30, 800, "Bsdk-log"))
    await page.evaluate(() => (globalThis as any).T.metric())
    run.mark("data-created")
    if (trigger === "flush") {
      await page.evaluate(() => (globalThis as any).T.flush())
      run.mark("flush-returned")
      await sleep(500)
    } else {
      await reload(page, run)
      await sleep(2000)
    }
    const spans = [...run.arrivedSpans().keys()].filter((n) => n.startsWith("Bsdk-span")).length
    const logs = [...run.arrivedLogs().keys()].filter((n) => n.startsWith("Bsdk-log")).length
    const metrics = run.records().filter((r) => r.method === "POST" && r.path.endsWith("/v1/metrics") && r.status === 200).length
    const reqs = run.pw.filter((p) => p.ev === "request" && /\/v1\/(traces|logs|metrics)$/.test(p.url)).map((p) => `${p.url.split("/").pop()}:${p.postBytes}`)
    const c = consoleCounts(run)
    const lost = spans < 40 || logs < 30
    return {
      verdict: lost ? "REPRODUCED (one signal rejected)" : "NOT REPRODUCED",
      details: `trigger=${trigger}; spans ${spans}/40, logs ${logs}/30, metrics POSTs ${metrics}; request bodies seen by Playwright [${reqs.join(" ")}]; console: ${[...c.flushFailed, ...c.flushSkipped].map((x) => x.text).join(" | ") || "none"}`,
      result: { spansDelivered: spans, logsDelivered: logs, metricsPosts: metrics, requestBodies: reqs },
    }
  })

// ---------------------------------------------------------------------------
// C: a single transient failure puts traces into a 60 s cooldown; the pagehide flush during
// that cooldown is skipped, so everything buffered (including the failed batch) is lost.
export const C = (browser: Browser, index: number, label: string, how: "reload" | "navigate" | "close", inject: boolean) =>
  withRun(browser, "C-unload-during-cooldown", label, index, { variant: "flush", interval: 1000 }, async (page, run) => {
    if (inject) addRulesFor(run, [{ pathIncludes: "/v1/traces", failStatus: 503, failCount: 1 }])
    await page.evaluate(() => (globalThis as any).T.span("C-first"))
    run.mark("C-first created")
    const firstPost = await waitSpanReceived(run, "C-first", 5000)
    run.mark("C-first POST received by server", { status: firstPost?.status })
    await sleep(1500)
    await page.evaluate(() => (globalThis as any).T.span("C-before-unload"))
    run.mark("C-before-unload created")
    await sleep(300)
    if (how === "reload") await reload(page, run)
    else if (how === "navigate") {
      run.mark("navigate-start")
      await page.goto(`http://localhost:${PAGE_PORT}/probe.html?tag=${run.tag}`)
      run.mark("navigate-done")
    } else {
      run.mark("close-start")
      await page.close({ runBeforeUnload: true })
      await sleep(200)
      run.mark("close-done")
    }
    await sleep(3000)
    const arrived = run.arrivedSpans()
    const c = consoleCounts(run)
    const first = arrived.has("C-first")
    const tail = arrived.has("C-before-unload")
    return {
      verdict: inject ? (!tail && !first ? "REPRODUCED (both spans lost)" : !tail ? "REPRODUCED (tail lost)" : "NOT REPRODUCED") : tail && first ? "CONTROL OK (both delivered)" : "CONTROL FAILED",
      details: `inject503=${inject} unload=${how}: C-first delivered=${first}, C-before-unload delivered=${tail}; console failed=${c.flushFailed.length} skipped=${c.flushSkipped.map((s) => `t=${s.t}`).join(",") || 0}`,
      result: { firstDelivered: first, tailDelivered: tail },
    }
  })

// ---------------------------------------------------------------------------
// D: the pagehide flush is serialized behind an in-flight periodic flush.
export const D = (browser: Browser, index: number, label: string, latencyMs: number, interval: number, tailBeforeReloadMs = 150) =>
  withRun(browser, "D-pagehide-behind-inflight-flush", label, index, { variant: "flush", interval }, async (page, run) => {
    // only the FIRST traces POST is slow (models one slow/hung round-trip through an ingest proxy)
    if (latencyMs > 0) addRulesFor(run, [{ pathIncludes: "/v1/traces", latencyMs, latencyCount: 1 }])
    await page.evaluate(() => (globalThis as any).T.span("D-early"))
    run.mark("D-early created")
    const early = await waitSpanReceived(run, "D-early", interval + 3000)
    run.mark("D-early POST received by server (response held for latencyMs)", { latencyMs })
    await sleep(100)
    await page.evaluate(() => (globalThis as any).T.span("D-tail"))
    run.mark("D-tail created", { msBeforeReload: tailBeforeReloadMs })
    await sleep(tailBeforeReloadMs)
    await reload(page, run)
    await sleep(latencyMs + 3000)
    const arrived = run.arrivedSpans()
    const tail = arrived.has("D-tail")
    const tailRec = run.records().find((r) => (r.summary as any)?.spanNames?.includes("D-tail"))
    const earlyRec = run.records().find((r) => (r.summary as any)?.spanNames?.includes("D-early"))
    return {
      verdict: latencyMs > 0 ? (tail ? "NOT REPRODUCED" : "REPRODUCED (tail span lost)") : tail ? "CONTROL OK" : "CONTROL FAILED",
      details: `first traces POST held ${latencyMs} ms, autoFlushInterval ${interval}, D-tail ended ${tailBeforeReloadMs} ms before reload: D-early delivered=${arrived.has("D-early")}${earlyRec ? ` (server responded t=${earlyRec.tResp !== undefined ? run.rel(earlyRec.tResp) : "?"}ms)` : ""}; D-tail delivered=${tail}${tailRec ? ` (received t=${run.rel(tailRec.tRecv)}ms)` : ""}; reload at t=${run.marks.find((m) => m.name === "reload-start")?.t}ms`,
      result: { earlyReceived: !!early, tailDelivered: tail },
    }
  })

// Dhang: head-of-line blocking — one slow traces POST stalls ALL later exports (traces, logs, metrics)
export const Dhang = (browser: Browser, index: number, label: string, holdMs: number) =>
  withRun(browser, "D-hang-head-of-line-blocking", label, index, { variant: "flush", interval: 1000 }, async (page, run) => {
    addRulesFor(run, [{ pathIncludes: "/v1/traces", latencyMs: holdMs, latencyCount: 1 }])
    await page.evaluate(() => (globalThis as any).T.span("Dh-first"))
    await waitSpanReceived(run, "Dh-first", 4000)
    const tHold = run.rel()
    run.mark("first traces POST received; server holds it", { holdMs })
    const created: { name: string; t: number }[] = []
    for (let i = 0; i < Math.floor(holdMs / 1000) + 3; i++) {
      await page.evaluate((i) => (globalThis as any).T.span(`Dh-${i}`), i)
      await page.evaluate((i) => (globalThis as any).T.logs(1, 0, `Dhlog-${i}`), i)
      created.push({ name: `Dh-${i}`, t: run.rel() })
      await sleep(1000)
    }
    await sleep(2000)
    const spans = run.arrivedSpans()
    const logs = run.arrivedLogs()
    const rows = created.map((c) => ({ name: c.name, createdAt: c.t, spanArrivedAt: spans.get(c.name), logArrivedAt: logs.get(`${c.name.replace("Dh-", "Dhlog-")}-0`) }))
    const delays = rows.filter((r) => r.spanArrivedAt !== undefined).map((r) => r.spanArrivedAt! - r.createdAt)
    const logDelays = rows.filter((r) => r.logArrivedAt !== undefined).map((r) => r.logArrivedAt! - r.createdAt)
    const traceReqs = run.records().filter((r) => r.method === "POST" && r.path.endsWith("/v1/traces"))
    return {
      verdict: Math.max(...delays) > holdMs / 2 ? "REPRODUCED (exports stalled behind one slow request)" : "NOT REPRODUCED",
      details: `first traces POST held ${holdMs} ms (from t=${tHold}ms): span export delay max ${Math.max(...delays)} ms / min ${Math.min(...delays)} ms; log export delay max ${Math.max(...logDelays)} ms (logs endpoint itself is fast); traces POSTs at server t=[${traceReqs.map((r) => `${run.rel(r.tRecv)}:${(r.summary as any).spanCount}sp/${r.bytes}B`).join(" ")}]`,
      result: { rows, holdMs },
    }
  })

// Dr: probabilistic version — steady span stream, reload at a random moment, realistic latency.
export const Dr = (browser: Browser, index: number, label: string, latencyMs: number, interval: number) =>
  withRun(browser, "Dr-random-reload-latency", label, index, { variant: "flush", interval }, async (page, run) => {
    if (latencyMs > 0) addRulesFor(run, [{ pathIncludes: "/v1/traces", latencyMs }])
    const created: { name: string; t: number }[] = []
    const until = Date.now() + 2000 + Math.floor(Math.random() * 4000)
    let k = 0
    while (Date.now() < until) {
      const name = `Dr-${k++}`
      await page.evaluate((n) => (globalThis as any).T.span(n), name)
      created.push({ name, t: run.rel() })
      await sleep(100)
    }
    const reloadAt = run.rel()
    const inflight = run.records().some((r) => r.path.endsWith("/v1/traces") && r.tResp === undefined)
    run.mark("reload decided", { traceFlushInFlightAtServer: inflight })
    await reload(page, run)
    await sleep(latencyMs + 3000)
    const arrived = run.arrivedSpans()
    const lost = created.filter((c) => !arrived.has(c.name))
    return {
      verdict: lost.length > 0 ? `LOSS (${lost.length} spans)` : "no loss",
      details: `latency ${latencyMs} ms, interval ${interval}: created ${created.length}, lost ${lost.length} (created ${lost.map((l) => reloadAt - l.t).join(",")} ms before reload); trace flush in flight at reload: ${inflight}`,
      result: { created: created.length, lost: lost.map((l) => ({ ...l, msBeforeReload: reloadAt - l.t })), inflightAtReload: inflight },
    }
  })

// ---------------------------------------------------------------------------
// E: combined unload keepalive budget — MapleFlush OTLP posts + replay meta/events/blob.
export const E = (browser: Browser, index: number, label: string, q: Q, opts: { spans: number; spanAttr: number; mutateChars: number; consoleLines: number }) =>
  withRun(browser, "E-combined-unload-budget", label, index, { variant: "flush", interval: 60000, ...q }, async (page, run) => {
    // let the replay engine start and ship its first full snapshot (chunk 0)
    if (q.replay !== 0) await run.waitForRecord((r) => (r.summary as any)?.kind === "replayBlob", 12_000)
    run.mark("first replay blob received (or replay disabled)")
    await sleep(200)
    if (opts.mutateChars > 0) await page.evaluate(([n]) => (globalThis as any).T.mutate(n, 20), [opts.mutateChars] as const)
    if (opts.consoleLines > 0) await page.evaluate((n) => { for (let i = 0; i < n; i++) console.log(`E-console-${i} ${"z".repeat(200)}`) }, opts.consoleLines)
    if (opts.spans > 0) await page.evaluate(([n, b]) => (globalThis as any).T.spans(n, b, "E-span"), [opts.spans, opts.spanAttr] as const)
    run.mark("payload created")
    await sleep(300)
    const reloadStart = run.rel()
    await reload(page, run)
    // wait long enough for the NEW page's first replay chunk (5 s timer) so we can see the chunk-seq gap
    if (q.replay !== 0) await run.waitForRecord((r) => (r.summary as any)?.kind === "replayBlob" && run.rel(r.tRecv) > reloadStart, 9000)
    await sleep(500)
    const unloadWindow = run.records().filter((r) => run.rel(r.tRecv) >= reloadStart - 50 && run.rel(r.tRecv) <= reloadStart + 1500 && r.method === "POST")
    const oldPage = {
      traces: unloadWindow.filter((r) => (r.summary as any)?.kind === "traces" && (r.summary as any).spanNames.some((n: string) => n.startsWith("E-span"))).map((r) => r.bytes),
      metaEnded: unloadWindow.filter((r) => (r.summary as any)?.kind === "sessionMeta" && (r.summary as any).rows.some((x: any) => x.status === "ended")).map((r) => r.bytes),
      sessionEvents: unloadWindow.filter((r) => (r.summary as any)?.kind === "sessionEvents").map((r) => r.bytes),
      replayBlob: unloadWindow.filter((r) => (r.summary as any)?.kind === "replayBlob").map((r) => r.bytes),
    }
    const chunkSeqs = run.records().filter((r) => (r.summary as any)?.kind === "replayBlob").map((r) => (r.summary as any).chunkSeq as number)
    const missingSeqs: number[] = []
    for (let i = 0; i <= Math.max(-1, ...chunkSeqs); i++) if (!chunkSeqs.includes(i)) missingSeqs.push(i)
    const pwUnload = run.pw.filter((p) => p.t >= reloadStart - 50 && p.t <= reloadStart + 1500)
    const c = consoleCounts(run)
    const spansDelivered = [...run.arrivedSpans().keys()].filter((n) => n.startsWith("E-span")).length
    const missing = [
      opts.spans > 0 && spansDelivered < opts.spans ? "traces" : null,
      oldPage.metaEnded.length === 0 ? "meta(ended)" : null,
      opts.consoleLines > 0 && oldPage.sessionEvents.length === 0 ? "final sessionEvents" : null,
      q.replay !== 0 && missingSeqs.length > 0 ? `replay chunk(s) ${missingSeqs.join(",")} (seq consumed, never posted)` : null,
    ].filter(Boolean)
    return {
      verdict: missing.length ? `LOST: ${missing.join(", ")}` : "all delivered",
      details: `spans ${spansDelivered}/${opts.spans}; old-page unload arrivals (bytes): traces[${oldPage.traces}] meta-ended[${oldPage.metaEnded}] events[${oldPage.sessionEvents}] blob[${oldPage.replayBlob}]; replay chunkSeqs at server [${chunkSeqs.join(",")}]; Playwright during unload: ${pwUnload.filter((p) => p.ev === "request").map((p) => `${p.url.split("/v1/")[1]}(${p.postBytes}B)`).join(" ")}`,
      result: { spansDelivered, oldPageUnloadArrivals: oldPage, chunkSeqs, missingSeqs, playwrightDuringUnload: pwUnload },
    }
  })

// ---------------------------------------------------------------------------
// F: replay/session posts that are NOT keepalive are aborted by navigation.
export const F = (browser: Browser, index: number, label: string, variant: string, target: "events" | "meta") =>
  withRun(browser, "F-replay-posts-aborted-by-navigation", label, index, { variant }, async (page, run) => {
    if (target === "events") {
      addRulesFor(run, [{ pathIncludes: "/v1/sessionEvents", latencyMs: 3000 }])
      // the replay engine (which installs the console capture) loads asynchronously, later with
      // Maple.layer than with MapleFlush.make; wait for its first "active" row before logging
      await run.waitForRecord((r) => (r.summary as any)?.kind === "sessionMeta", 5000)
      await sleep(1000)
      await page.evaluate(() => {
        for (let i = 0; i < 5; i++) console.log(`F-console-${i}`)
      })
      // the event sink flushes every 5 s with keepalive=false; wait until that POST is at the server
      const rec = await run.waitForRecord((r) => (r.summary as any)?.kind === "sessionEvents" && (r.summary as any).consoleMessages.some((m: string) => m.startsWith("F-console")), 8000)
      run.mark("periodic sessionEvents POST received by server; response held 3 s", { found: !!rec })
      await sleep(100)
      await reload(page, run)
    } else {
      addRulesFor(run, [{ pathIncludes: "/v1/sessionReplays/meta", latencyMs: 3000 }])
      // the new page posts an "active" meta row (keepalive=false) right at startup; reload while it is in flight
      await reload(page, run)
      await run.waitForRecord((r) => (r.summary as any)?.kind === "sessionMeta" && run.rel(r.tRecv) > (run.marks.find((m) => m.name === "reload-start")?.t ?? 0), 5000)
      await sleep(300)
      await reload(page, run)
    }
    await sleep(3500)
    const c = consoleCounts(run)
    const failedPw = run.pw.filter((p) => p.ev === "failed")
    return {
      verdict: target === "events" && !run.marks.find((m) => m.name.startsWith("periodic sessionEvents POST received"))?.found ? "PRECONDITION NOT MET (no in-flight events POST at reload)" : c.replayPostFailed.length > 0 ? "REPRODUCED" : "NOT REPRODUCED",
      details: `${variant}/${target}: console: ${c.replayPostFailed.map((x) => x.text).join(" | ") || "none"}; Playwright requestfailed: ${failedPw.map((p) => `${p.url.split("/v1/")[1]}[${p.failure}]`).join(" ") || "none"}`,
      result: { replayPostFailed: c.replayPostFailed, playwrightFailed: failedPw },
    }
  })

// Fr: realistic — modest latency on every ingest path, several reloads at random moments.
export const Fr = (browser: Browser, index: number, label: string, variant: string, latencyMs: number) =>
  withRun(browser, "Fr-random-reloads", label, index, { variant, interval: 1000 }, async (page, run) => {
    addRulesFor(run, [{ pathIncludes: "/otlp/", latencyMs }])
    for (let i = 0; i < 8; i++) {
      await page.evaluate((i) => {
        console.log(`Fr-console-${i}`)
        document.getElementById("btn-spans")?.dispatchEvent(new MouseEvent("click", { bubbles: true }))
      }, i)
      await sleep(500 + Math.floor(Math.random() * 6500))
      await reload(page, run)
    }
    await sleep(latencyMs + 3000)
    const c = consoleCounts(run)
    const failedPw = run.pw.filter((p) => p.ev === "failed")
    const byPath: Record<string, number> = {}
    for (const p of failedPw) byPath[`${p.url.split("/v1/")[1]}[${p.failure}]`] = (byPath[`${p.url.split("/v1/")[1]}[${p.failure}]`] ?? 0) + 1
    return {
      verdict: c.replayPostFailed.length > 0 || c.flushFailed.length > 0 ? "REPRODUCED" : "NOT REPRODUCED",
      details: `${variant}, latency ${latencyMs} ms, 8 reloads: replay-post console errors ${c.replayPostFailed.length} [${[...new Set(c.replayPostFailed.map((x) => x.text.split(" failed")[0]))].join("; ")}], SDK flush failed ${c.flushFailed.length}, skipped ${c.flushSkipped.length}; Playwright requestfailed by path: ${JSON.stringify(byPath)}`,
      result: { replayPostFailed: c.replayPostFailed, flushFailed: c.flushFailed, flushSkipped: c.flushSkipped, playwrightFailedByPath: byPath },
    }
  })

// ---------------------------------------------------------------------------
// G: is Playwright's requestfailed/ERR_ABORTED trustworthy for keepalive requests?
export const G = (browser: Browser, index: number, label: string, withReload: boolean) =>
  withRun(browser, "G-playwright-keepalive-artifact", label, index, { variant: "flush", interval: 1000 }, async (page, run) => {
    for (let i = 0; i < 20; i++) {
      await page.evaluate((i) => (globalThis as any).T.span(`G-${i}`), i)
      await sleep(400)
    }
    if (withReload) {
      await page.evaluate(() => (globalThis as any).T.span("G-tail"))
      await sleep(150)
      await reload(page, run)
    }
    await sleep(3000)
    // join Playwright's view with the server's view, per request
    const posts = run.pw.filter((p) => p.ev === "request" && p.method === "POST")
    const fin = run.pw.filter((p) => p.ev === "finished" && p.method === "POST")
    const fail = run.pw.filter((p) => p.ev === "failed" && p.method === "POST")
    const server200 = run.records().filter((r) => r.method === "POST" && r.status === 200)
    const kinds = ["traces", "sessionReplays/meta", "sessionEvents", "sessionReplays/blob", "logs", "metrics"]
    const table: Record<string, { pwRequest: number; pwFinished: number; pwFailed: number; failures: string[]; server200: number }> = {}
    for (const k of kinds) {
      const m = (u: string) => u.endsWith(`/v1/${k}`)
      table[k] = {
        pwRequest: posts.filter((p) => m(p.url)).length,
        pwFinished: fin.filter((p) => m(p.url)).length,
        pwFailed: fail.filter((p) => m(p.url)).length,
        failures: [...new Set(fail.filter((p) => m(p.url)).map((p) => p.failure ?? ""))],
        server200: server200.filter((r) => m(r.path)).length,
      }
    }
    const tracesFailedButDelivered = table.traces!.pwFailed > 0 && table.traces!.server200 >= table.traces!.pwRequest
    return {
      verdict: fail.length > 0 && server200.length >= posts.length ? "ARTIFACT (Playwright reports failures for requests the server answered 200)" : fail.length > 0 ? "MIXED" : "no requestfailed events",
      details: `reload=${withReload}: ${Object.entries(table).map(([k, v]) => `${k}: pw req ${v.pwRequest}/fin ${v.pwFinished}/fail ${v.pwFailed}${v.failures.length ? `[${v.failures.join(",")}]` : ""} vs server200 ${v.server200}`).join("; ")}`,
      result: { table, tracesFailedButDelivered },
    }
  })

// ---------------------------------------------------------------------------
// H: Maple.layer has no unload flush (documented) — baseline + a user-land workaround.
export const H = (browser: Browser, index: number, label: string, variant: string) =>
  withRun(browser, "H-maple-layer-unload", label, index, { variant }, async (page, run) => {
    await page.evaluate(() => (globalThis as any).T.span("H-early"))
    run.mark("H-early created")
    await sleep(6500)
    await page.evaluate(() => (globalThis as any).T.span("H-tail"))
    run.mark("H-tail created")
    await sleep(1000)
    await reload(page, run)
    await sleep(3000)
    const arrived = run.arrivedSpans()
    return {
      verdict: arrived.has("H-tail") ? "tail delivered" : "tail LOST",
      details: `${variant}: H-early (created 7.5 s before reload) delivered=${arrived.has("H-early")}; H-tail (created 1 s before reload) delivered=${arrived.has("H-tail")}`,
      result: { early: arrived.has("H-early"), tail: arrived.has("H-tail") },
    }
  })

// ---------------------------------------------------------------------------
// I-cross: cross-origin endpoint (CORS preflight) + pagehide flush; also records request headers.
export const Icross = (browser: Browser, index: number, label: string, origin: "same" | "cross") =>
  withRun(browser, "I-cross-origin", label, index, { variant: "flush", origin, interval: 5000 }, async (page, run) => {
    await page.evaluate(() => (globalThis as any).T.span("I-tail"))
    await sleep(150)
    await reload(page, run)
    await sleep(2500)
    const arrived = run.arrivedSpans()
    const pre = run.records().filter((r) => r.method === "OPTIONS")
    const tracePost = run.records().find((r) => r.method === "POST" && r.path.endsWith("/v1/traces"))
    return {
      verdict: arrived.has("I-tail") ? "tail delivered" : "tail LOST",
      details: `${origin}-origin: I-tail delivered=${arrived.has("I-tail")}; preflights ${pre.length} (access-control-request-headers: ${[...new Set(pre.map((p) => p.headers["access-control-request-headers"]))].join(" | ") || "n/a"}); traces POST user-agent header: ${tracePost?.headers["user-agent"] ?? "n/a"}, x-maple-sdk: ${tracePost?.headers["x-maple-sdk"] ?? "n/a"}`,
      result: { tail: arrived.has("I-tail"), preflightRequestHeaders: pre.map((p) => p.headers["access-control-request-headers"]), tracePostHeaders: tracePost?.headers },
    }
  })

// ---------------------------------------------------------------------------
// K: (follow-up to a real-app observation) action = 1 span + 1 Effect log, then reload ~150 ms later,
// autoFlushInterval 1000. Which document / which flush prints "traces|logs flush failed", and was the
// data actually delivered?
export const K = (browser: Browser, index: number, label: string, q: Q) =>
  withRun(browser, "K-reload-150ms-after-action", label, index, { variant: "flush", interval: 1000, ...q }, async (page, run) => {
    await sleep(1500)
    await page.evaluate(() => (globalThis as any).T.span("K-action-span"))
    await page.evaluate(() => (globalThis as any).T.logs(1, 0, "K-action-log"))
    run.mark("action (span + log) done")
    await sleep(150)
    await reload(page, run)
    await sleep(3000)
    const span = run.arrivedSpans().has("K-action-span")
    const log = run.arrivedLogs().has("K-action-log-0")
    const c = consoleCounts(run)
    const reloadAt = run.marks.find((m) => m.name === "reload-start")!.t
    const pw = run.pw.filter((p) => p.t >= reloadAt - 20 && p.t <= reloadAt + 500)
    const srv = run.records().filter((r) => r.method === "POST" && run.rel(r.tRecv) >= reloadAt - 20 && run.rel(r.tRecv) <= reloadAt + 500)
    return {
      verdict: `span ${span ? "delivered" : "LOST"}, log ${log ? "delivered" : "LOST"}; console flush-failed: [${c.flushFailed.map((f) => f.text.split(";")[0]!.replace("[MapleClientSDK] ", "")).join(", ")}]`,
      details: `reload at t=${reloadAt}; requests issued around the reload (Playwright): ${pw.filter((p) => p.ev === "request").map((p) => `${p.url.split("/v1/")[1]}(${p.postBytes}B)@${p.t}`).join(" ")}; Playwright failed: ${pw.filter((p) => p.ev === "failed").map((p) => `${p.url.split("/v1/")[1]}[${p.failure}]`).join(" ") || "none"}; server 200 in same window: ${srv.map((r) => `${r.path.split("/v1/")[1]}@${run.rel(r.tRecv)}`).join(" ")}; console: ${[...c.flushFailed, ...c.flushSkipped, ...c.replayPostFailed].map((x) => `${x.text.split(":")[0]}@${x.t}`).join(" | ")}`,
      result: { spanDelivered: span, logDelivered: log, playwrightAroundReload: pw },
    }
  })

// L: multi-argument Effect log -> OTLP log body encoding
export const L = (browser: Browser, index: number) =>
  withRun(browser, "L-multi-arg-log-body", "logError-4-args", index, { variant: "flush", interval: 1000 }, async (page, run) => {
    await page.evaluate(() => (globalThis as any).T.logMulti("L"))
    await sleep(2500)
    const rec = run.records().find((r) => (r.summary as any)?.kind === "logs")
    const raw = (rec?.summary as any)?.firstRawBodies?.[0]
    return {
      verdict: raw?.arrayValue ? "body is an OTLP arrayValue (structured), not a JSON string" : raw?.stringValue ? "body is a stringValue" : "no log arrived",
      details: `OTLP logRecord.body as POSTed by the SDK: ${JSON.stringify(raw)}`,
      result: { rawBody: raw },
    }
  })

// W: user-land workaround check — A-style burst with the ?shim=1 fetch wrapper (downgrade oversized keepalive POSTs)
export const W = (browser: Browser, index: number, label: string, shim: boolean) =>
  withRun(browser, "W-workaround-keepalive-downgrade-shim", label, index, { variant: "flush", interval: 1000, ...(shim ? { shim: 1 } : {}) }, async (page, run) => {
    await sleep(1000)
    await page.evaluate(() => (globalThis as any).T.spans(200, 500, "W-burst"))
    for (let k = 0; k < 5; k++) {
      await sleep(1000)
      await page.evaluate((k) => (globalThis as any).T.span(`W-tick-${k}`), k)
    }
    await sleep(2500)
    const arrived = [...run.arrivedSpans().keys()]
    const burst = arrived.filter((n) => n.startsWith("W-burst")).length
    const ticks = arrived.filter((n) => n.startsWith("W-tick")).length
    const c = consoleCounts(run)
    return {
      verdict: burst === 200 && ticks === 5 ? "all delivered" : `LOSS (burst ${burst}/200, ticks ${ticks}/5)`,
      details: `shim=${shim}: burst spans delivered ${burst}/200, later spans ${ticks}/5; console flush-failed ${c.flushFailed.length}, skipped ${c.flushSkipped.length}`,
      result: { burst, ticks },
    }
  })
@@@@@@@@@@ END FILE: repro/scenarios/sdk.ts

@@@@@@@@@@ FILE: repro/server.ts (13999 bytes)
/**
 * Mock Maple / OTLP ingest + static file server for the keepalive repro.
 *
 * Two listeners:
 *   - PAGE_PORT  (default 4173): serves the test pages (/ , /probe.html, /dist/*)
 *                                AND a same-origin ingest under /otlp/*
 *   - CROSS_PORT (default 4174): serves ONLY the ingest under /otlp/* (cross-origin
 *                                from the page, so the browser must CORS-preflight).
 *
 * Every ingest request (including OPTIONS preflights) is recorded in memory and
 * appended to a JSONL log. Each record carries: wall-clock timestamps for
 * "request headers received", "body fully read" and "response sent", the path,
 * body byte size, interesting headers, and a parsed summary of the payload
 * (span names, log bodies, metric names, session-event types/seqs, replay
 * chunk headers).
 *
 * Failure / latency injection is driven by "rules" (see `Rule` below), set via
 * POST /__control (JSON) or programmatically through `startServer().setRules()`.
 *
 * Run standalone:  bun run server.ts      (then open http://localhost:4173/)
 */
import { appendFileSync, existsSync, mkdirSync, readFileSync } from "node:fs"
import { dirname, join } from "node:path"

export interface Rule {
  /** Substring match on the request path, e.g. "/v1/traces". */
  readonly pathIncludes: string
  /** Optional run tag; rule only applies to requests of that run (tag = path segment after /otlp/). */
  readonly tag?: string
  /** Delay before responding (ms). The body is read first, then we wait. */
  readonly latencyMs?: number
  /** Apply `latencyMs` only to the first N matching requests (default: all). */
  readonly latencyCount?: number
  /** Respond with this status instead of 200 ... */
  readonly failStatus?: number
  /** ... for the first N matching requests only (default: all). */
  readonly failCount?: number
}

export interface IngestRecord {
  readonly seq: number
  readonly port: number
  readonly method: string
  readonly path: string
  readonly query: string
  readonly tag: string | null
  readonly origin: string | null
  readonly tRecv: number
  tBody?: number
  tResp?: number
  bytes?: number
  status?: number
  clientAborted?: boolean
  readonly headers: Record<string, string>
  summary?: unknown
  latencyMs?: number
  injectedFailure?: boolean
}

const ROOT = dirname(new URL(import.meta.url).pathname)
const PUBLIC = join(ROOT, "public")

const INTERESTING_HEADERS = [
  "content-type",
  "content-length",
  "authorization",
  "x-maple-sdk",
  "user-agent",
  "origin",
  "referer",
  "sec-fetch-mode",
  "sec-fetch-dest",
  "sec-fetch-site",
  "access-control-request-method",
  "access-control-request-headers",
  "x-maple-session-id",
  "x-maple-chunk-seq",
  "x-maple-is-checkpoint",
  "x-maple-event-count",
  "x-maple-duration-ms",
]

const pickHeaders = (req: Request): Record<string, string> => {
  const out: Record<string, string> = {}
  for (const name of INTERESTING_HEADERS) {
    const v = req.headers.get(name)
    if (v === null) continue
    // never log real credentials; the repro only ever sends the "~" placeholder
    out[name] = name === "authorization" ? (v === "Bearer ~" ? v : "Bearer <redacted>") : name === "user-agent" ? v.slice(0, 60) : v
  }
  return out
}

const tagFrom = (req: Request, url: URL): string | null => {
  // The page configures the SDK endpoint as `<origin>/otlp/<tag>`, so every SDK
  // request path is `/otlp/<tag>/v1/...` — this works cross-origin too, where the
  // Referer is reduced to the bare origin.
  const m = /^\/otlp\/([^/]+)\/v1\//.exec(url.pathname)
  if (m) return decodeURIComponent(m[1]!)
  const direct = url.searchParams.get("tag")
  if (direct) return direct
  const ref = req.headers.get("referer")
  if (!ref) return null
  try {
    return new URL(ref).searchParams.get("tag")
  } catch {
    return null
  }
}

const anyValueToString = (v: any): string => {
  if (!v || typeof v !== "object") return String(v)
  if ("stringValue" in v) return v.stringValue
  if ("intValue" in v) return String(v.intValue)
  if ("doubleValue" in v) return String(v.doubleValue)
  if ("boolValue" in v) return String(v.boolValue)
  return JSON.stringify(v).slice(0, 200)
}

const summarize = (path: string, req: Request, body: Uint8Array): unknown => {
  try {
    if (path.endsWith("/v1/traces")) {
      const json = JSON.parse(new TextDecoder().decode(body))
      const names: string[] = []
      for (const rs of json.resourceSpans ?? []) for (const ss of rs.scopeSpans ?? []) for (const s of ss.spans ?? []) names.push(s.name)
      return { kind: "traces", spanCount: names.length, spanNames: names }
    }
    if (path.endsWith("/v1/logs")) {
      const json = JSON.parse(new TextDecoder().decode(body))
      const bodies: string[] = []
      const rawBodies: unknown[] = []
      for (const rl of json.resourceLogs ?? []) for (const sl of rl.scopeLogs ?? []) for (const r of sl.logRecords ?? []) {
        bodies.push(anyValueToString(r.body).slice(0, 80))
        if (rawBodies.length < 3) rawBodies.push(r.body)
      }
      return { kind: "logs", logCount: bodies.length, logBodies: bodies, firstRawBodies: rawBodies }
    }
    if (path.endsWith("/v1/metrics")) {
      const json = JSON.parse(new TextDecoder().decode(body))
      const names: string[] = []
      for (const rm of json.resourceMetrics ?? []) for (const sm of rm.scopeMetrics ?? []) for (const m of sm.metrics ?? []) names.push(m.name)
      return { kind: "metrics", metricCount: names.length, metricNames: names }
    }
    if (path.endsWith("/v1/sessionEvents")) {
      const rows = new TextDecoder().decode(body).split("\n").filter(Boolean).map((l) => JSON.parse(l))
      return {
        kind: "sessionEvents",
        rowCount: rows.length,
        sessionId: rows[0]?.session_id,
        seqs: rows.map((r) => r.seq),
        types: rows.map((r) => (r.type === "console" ? `console.${r.level}` : r.type)),
        consoleMessages: rows.filter((r) => r.type === "console").map((r) => String(r.message).slice(0, 160)),
      }
    }
    if (path.endsWith("/v1/sessionReplays/meta")) {
      const rows = new TextDecoder().decode(body).split("\n").filter(Boolean).map((l) => JSON.parse(l))
      return {
        kind: "sessionMeta",
        rows: rows.map((r) => ({ session_id: r.session_id, status: r.status, version: r.version, trace_ids: r.trace_ids?.length })),
      }
    }
    if (path.endsWith("/v1/sessionReplays/blob")) {
      return {
        kind: "replayBlob",
        sessionId: req.headers.get("x-maple-session-id"),
        chunkSeq: Number(req.headers.get("x-maple-chunk-seq")),
        isCheckpoint: req.headers.get("x-maple-is-checkpoint"),
        eventCount: Number(req.headers.get("x-maple-event-count")),
        gzippedBytes: body.byteLength,
      }
    }
    if (path.startsWith("/probe/")) {
      return { kind: "probe", bytes: body.byteLength }
    }
    return { kind: "unknown", bytes: body.byteLength }
  } catch (error) {
    return { kind: "unparseable", error: String(error) }
  }
}

const corsHeaders = (req: Request): Record<string, string> => {
  const origin = req.headers.get("origin")
  const h: Record<string, string> = {
    "access-control-allow-methods": "POST, OPTIONS",
    "access-control-max-age": "0", // force a preflight for every cross-origin request (worst case)
    vary: "origin",
  }
  if (origin) {
    h["access-control-allow-origin"] = origin
    h["access-control-allow-credentials"] = "true"
  }
  const requested = req.headers.get("access-control-request-headers")
  if (requested) h["access-control-allow-headers"] = requested
  return h
}

const sleep = (ms: number) => new Promise((r) => setTimeout(r, ms))

const contentTypeOf = (p: string) =>
  p.endsWith(".html") ? "text/html; charset=utf-8" : p.endsWith(".js") ? "text/javascript; charset=utf-8" : p.endsWith(".map") ? "application/json" : "application/octet-stream"

export interface ServerHandle {
  readonly pagePort: number
  readonly crossPort: number
  readonly records: IngestRecord[]
  setRules(rules: Rule[]): void
  addRules(rules: Rule[]): void
  clearRulesForTag(tag: string): void
  reset(): void
  stop(): void
}

export const startServer = (opts: { pagePort?: number; crossPort?: number; logFile?: string } = {}): ServerHandle => {
  const pagePort = opts.pagePort ?? Number(process.env.PAGE_PORT ?? 4173)
  const crossPort = opts.crossPort ?? Number(process.env.CROSS_PORT ?? 4174)
  const logFile = opts.logFile
  if (logFile) mkdirSync(dirname(logFile), { recursive: true })
  const records: IngestRecord[] = []
  let rules: Rule[] = []
  const ruleHits = new Map<Rule, number>()
  const latencyHits = new Map<Rule, number>()
  let seq = 0

  const persist = (rec: IngestRecord) => {
    if (logFile) appendFileSync(logFile, `${JSON.stringify(rec)}\n`)
  }

  const handleIngest = async (req: Request, url: URL, port: number): Promise<Response> => {
    const rec: IngestRecord = {
      seq: seq++,
      port,
      method: req.method,
      path: url.pathname,
      query: url.search,
      tag: tagFrom(req, url),
      origin: req.headers.get("origin"),
      tRecv: Date.now(),
      headers: pickHeaders(req),
    }
    records.push(rec)
    const cors = corsHeaders(req)
    if (req.method === "OPTIONS") {
      rec.status = 204
      rec.tResp = Date.now()
      persist(rec)
      return new Response(null, { status: 204, headers: cors })
    }
    let body: Uint8Array
    try {
      body = new Uint8Array(await req.arrayBuffer())
    } catch (error) {
      rec.clientAborted = true
      rec.summary = { kind: "body-read-failed", error: String(error) }
      rec.tResp = Date.now()
      persist(rec)
      return new Response(null, { status: 499, headers: cors })
    }
    rec.tBody = Date.now()
    rec.bytes = body.byteLength
    rec.summary = summarize(url.pathname, req, body)

    // probe endpoints take their latency from the query string
    let latency = Number(url.searchParams.get("delay") ?? 0)
    let failStatus: number | undefined
    for (const rule of rules) {
      if (!url.pathname.includes(rule.pathIncludes)) continue
      if (rule.tag !== undefined && rule.tag !== rec.tag) continue
      if (rule.latencyMs) {
        const hits = latencyHits.get(rule) ?? 0
        if (rule.latencyCount === undefined || hits < rule.latencyCount) {
          latencyHits.set(rule, hits + 1)
          latency = Math.max(latency, rule.latencyMs)
        }
      }
      if (rule.failStatus) {
        const hits = ruleHits.get(rule) ?? 0
        if (rule.failCount === undefined || hits < rule.failCount) {
          ruleHits.set(rule, hits + 1)
          failStatus = rule.failStatus
        }
      }
    }
    if (latency > 0) {
      rec.latencyMs = latency
      let aborted = false
      req.signal?.addEventListener?.("abort", () => {
        aborted = true
      })
      await sleep(latency)
      if (aborted) rec.clientAborted = true
    }
    const status = failStatus ?? 200
    rec.injectedFailure = failStatus !== undefined
    rec.status = status
    rec.tResp = Date.now()
    persist(rec)
    return new Response(status === 200 ? "{}" : `injected ${status}`, {
      status,
      headers: { ...cors, "content-type": "application/json" },
    })
  }

  const handleControl = async (req: Request, url: URL): Promise<Response> => {
    if (url.pathname === "/__control" && req.method === "POST") {
      const json = (await req.json()) as { rules?: Rule[] }
      rules = json.rules ?? []
      ruleHits.clear()
      return Response.json({ ok: true, rules })
    }
    if (url.pathname === "/__reset") {
      records.length = 0
      rules = []
      ruleHits.clear()
      return Response.json({ ok: true })
    }
    if (url.pathname === "/__log") {
      const since = Number(url.searchParams.get("since") ?? 0)
      const tag = url.searchParams.get("tag")
      return Response.json(records.filter((r) => r.seq >= since && (tag === null || r.tag === tag)))
    }
    return new Response("not found", { status: 404 })
  }

  const serveStatic = (url: URL): Response => {
    let p = url.pathname
    if (p === "/" || p === "/index.html") p = "/index.html"
    const file = join(PUBLIC, p)
    if (!file.startsWith(PUBLIC) || !existsSync(file)) return new Response("not found", { status: 404 })
    return new Response(readFileSync(file), {
      headers: { "content-type": contentTypeOf(p), "cache-control": "no-store" },
    })
  }

  const page = Bun.serve({
    port: pagePort,
    idleTimeout: 60,
    async fetch(req) {
      const url = new URL(req.url)
      if (url.pathname.startsWith("/otlp/") || url.pathname.startsWith("/probe/")) return handleIngest(req, url, pagePort)
      if (url.pathname.startsWith("/__")) return handleControl(req, url)
      return serveStatic(url)
    },
  })
  const cross = Bun.serve({
    port: crossPort,
    idleTimeout: 60,
    async fetch(req) {
      const url = new URL(req.url)
      if (url.pathname.startsWith("/otlp/") || url.pathname.startsWith("/probe/")) return handleIngest(req, url, crossPort)
      return new Response("not found", { status: 404, headers: corsHeaders(req) })
    },
  })

  return {
    pagePort,
    crossPort,
    records,
    setRules(next) {
      rules = next
      ruleHits.clear()
    },
    addRules(next) {
      rules = [...rules, ...next]
    },
    clearRulesForTag(tag) {
      rules = rules.filter((r) => r.tag !== tag)
    },
    reset() {
      records.length = 0
      rules = []
      ruleHits.clear()
    },
    stop() {
      page.stop(true)
      cross.stop(true)
    },
  }
}

if (import.meta.main) {
  const logFile = join(ROOT, "server-requests.jsonl")
  const s = startServer({ logFile })
  console.log(`page + same-origin ingest: http://localhost:${s.pagePort}/`)
  console.log(`cross-origin ingest:       http://127.0.0.1:${s.crossPort}/otlp`)
  console.log(`request log:               ${logFile}`)
  console.log(`try: http://localhost:${s.pagePort}/?variant=flush&interval=1000&tag=manual`)
}
@@@@@@@@@@ END FILE: repro/server.ts

@@@@@@@@@@ FILE: repro/variants/sdk-0.10.0/bun.lock (798 bytes)
{
  "lockfileVersion": 2,
  "configVersion": 1,
  "workspaces": {
    "": {
      "name": "maple-repro-variant-sdk-0-10-0",
      "dependencies": {
        "@maple-dev/effect-sdk": "0.10.0",
        "effect": "4.0.0",
      },
    },
  },
  "packages": {
    "@maple-dev/effect-sdk": ["@maple-dev/effect-sdk@0.10.0", "", { "dependencies": { "std-env": "^4.0.0" }, "peerDependencies": { "effect": "^4.0.0" } }, "sha512-3MWyBqWPyiizAZItDk7Uk+vi69csziIJxCcaqW10aowVV1Tlcwao1EW6xLccAzvfkz43Y8WAiLzhf/Kr4oRkdQ=="],

    "effect": ["effect@4.0.0", "", {}, "sha512-ooc1TG5t+FfzgYnFz2ff6BBKyZ7EwBRVXC7c4RhQUAD6/TZ2gTXXMeb4WX7a19ozQo4J73/QW+S00YAIresoMQ=="],

    "std-env": ["std-env@4.3.0", "", {}, "sha512-OtU/EgQ1kIm5KwqQpBC6ZEMXrZRui11w8zgfTWp8cdO9B8OaPsbA8bTHO2P+HNo1VlUTGMVBwPhydu6poeXiag=="],
  }
}
@@@@@@@@@@ END FILE: repro/variants/sdk-0.10.0/bun.lock

@@@@@@@@@@ FILE: repro/variants/sdk-0.10.0/package.json (415 bytes)
{
  "name": "maple-repro-variant-sdk-0-10-0",
  "private": true,
  "type": "module",
  "description": "Optional: the same test page built against the latest published SDK (0.10.0) and its Effect peer (4.0.0), to check which findings persist. Built by `bun run build.ts` in the parent folder when this folder has node_modules.",
  "dependencies": {
    "@maple-dev/effect-sdk": "0.10.0",
    "effect": "4.0.0"
  }
}
@@@@@@@@@@ END FILE: repro/variants/sdk-0.10.0/package.json

@@@@@@@@@@ FILE: sdk-excerpts/01-flush-core-cooldown-serialization.txt (7677 bytes)
Flush core: cooldown, flushSignal, guardFlush, makeSerializedFlush, runFlush
============================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/version-B225m7RW.mjs  lines 1-173  (src/shared/flush-core.ts region) ---
     1	import { Array as Array$1, Cause, Context, Effect, Exit, Layer, Logger, Metric, Option, Predicate, Redacted, References, Result, Tracer } from "effect";
     2	import * as OtlpResource from "effect/unstable/observability/OtlpResource";
     3	import * as ErrorReporter from "effect/ErrorReporter";
     4	//#region src/shared/try-sync.ts
     5	/** Run a throwing synchronous call, `None` if it threw. */
     6	const trySync = (thunk) => Effect.runSync(Effect.option(Effect.try(thunk)));
     7	/** Run a throwing synchronous call, `undefined` if it threw. */
     8	const trySyncOrUndefined = (thunk) => Option.getOrUndefined(trySync(thunk));
     9	//#endregion
    10	//#region src/shared/flush-core.ts
    11	/** Disable a signal for this long after a failed POST so a broken collector isn't hammered. */
    12	const COOLDOWN_MS = 6e4;
    13	/**
    14	* Turn a resolved resource into ready-to-POST URLs + headers. Shared by all
    15	* presets; `userAgent` is the only per-preset difference.
    16	*/
    17	const buildResolved = (r, opts) => {
    18		const base = r.endpoint ?? "https://ingest.maple.dev";
    19		const baseUrl = base.endsWith("/") ? base.slice(0, -1) : base;
    20		const tracesUrl = `${baseUrl}${opts.tracesPath ?? "/v1/traces"}`;
    21		const logsUrl = `${baseUrl}${opts.logsPath ?? "/v1/logs"}`;
    22		const metricsUrl = `${baseUrl}${opts.metricsPath ?? "/v1/metrics"}`;
    23		const headers = {
    24			"content-type": "application/json",
    25			"user-agent": opts.userAgent,
    26			"x-maple-sdk": opts.userAgent
    27		};
    28		if (r.ingestKey) headers.authorization = `Bearer ${Redacted.value(r.ingestKey)}`;
    29		return {
    30			tracesUrl,
    31			logsUrl,
    32			metricsUrl,
    33			resource: makeOtlpResource(r.resource),
    34			scope: { name: r.resource.serviceName },
    35			headers,
    36			noOp: r.ingestKey === void 0
    37		};
    38	};
    39	const makeOtlpResource = (resource) => {
    40		const attrs = [];
    41		for (const [key, value] of Object.entries(resource.attributes)) attrs.push({
    42			key,
    43			value: anyValue(value)
    44		});
    45		attrs.push({
    46			key: "service.name",
    47			value: { stringValue: resource.serviceName }
    48		});
    49		if (resource.serviceVersion) attrs.push({
    50			key: "service.version",
    51			value: { stringValue: resource.serviceVersion }
    52		});
    53		return {
    54			attributes: attrs,
    55			droppedAttributesCount: 0
    56		};
    57	};
    58	const anyValue = (value) => {
    59		if (Array.isArray(value)) return { arrayValue: { values: value.map(anyValue) } };
    60		switch (typeof value) {
    61			case "string": return { stringValue: value };
    62			case "boolean": return { boolValue: value };
    63			case "number": return Number.isInteger(value) ? { intValue: value } : { doubleValue: value };
    64			case "bigint": return { intValue: Number(value) };
    65			default: return { stringValue: String(value) };
    66		}
    67	};
    68	/** Plain `fetch` POST. Throws on non-2xx so {@link flushSignal} records a cooldown. */
    69	const post = async (url, headers, body) => {
    70		const res = await fetch(url, {
    71			method: "POST",
    72			headers,
    73			body: JSON.stringify(body)
    74		});
    75		if (!res.ok) throw new Error(`OTLP ${res.status} ${res.statusText}`);
    76	};
    77	/** Default transport: plain `fetch` (server + Cloudflare). */
    78	const fetchTransport = { post };
    79	const flushSignal = async (args) => {
    80		const { url, headers, buffer, body, state, signal, transport, logPrefix } = args;
    81		if (state.disabledUntil && Date.now() < state.disabledUntil) {
    82			console.warn(`${logPrefix} ${signal} flush skipped (cooldown ${state.disabledUntil - Date.now()}ms remaining)`);
    83			return;
    84		}
    85		state.disabledUntil = 0;
    86		const batch = buffer.drain();
    87		if (batch.length === 0) return;
    88		const posted = await Effect.runPromise(Effect.result(Effect.tryPromise({
    89			try: () => transport.post(url, headers, body(batch)),
    90			catch: (cause) => cause
    91		})));
    92		if (Result.isFailure(posted)) {
    93			buffer.restore(batch);
    94			state.disabledUntil = Date.now() + COOLDOWN_MS;
    95			console.error(`${logPrefix} ${signal} flush failed; cooldown 60s:`, posted.failure);
    96		}
    97	};
    98	/**
    99	* Wrap a flush body so it logs its failure rather than rejecting.
   100	*
   101	* Every flush is documented never to reject: they are fired as `void flush()`
   102	* from auto-flush timers, `pagehide`/`visibilitychange` handlers, and
   103	* `ctx.waitUntil`, where a rejection is an unhandled rejection — fatal under
   104	* `--unhandled-rejections=strict` — caused purely by telemetry. `runFlush`
   105	* already absorbs per-signal transport errors; this covers everything around it,
   106	* chiefly resource resolution.
   107	*/
   108	const guardFlush = (logPrefix, run) => async (...args) => {
   109		const outcome = await Effect.runPromise(Effect.result(Effect.tryPromise({
   110			try: () => run(...args),
   111			catch: (cause) => cause
   112		})));
   113		if (Result.isFailure(outcome)) console.error(`${logPrefix} flush failed:`, outcome.failure);
   114	};
   115	/** Serialize flush calls so concurrent timers/manual hooks cannot drain overlapping batches. */
   116	const makeSerializedFlush = (run) => {
   117		let tail = Promise.resolve();
   118		return (...args) => {
   119			const next = tail.then(() => run(...args), () => run(...args));
   120			tail = next.catch(() => void 0);
   121			return next;
   122		};
   123	};
   124	/**
   125	* Drain the span + log buffers and POST them. Errors are swallowed per signal
   126	* (see {@link flushSignal}); the returned promise never rejects.
   127	*
   128	* - `noOp`: drain so the buffers don't grow unbounded, fire `onNoOp` (one-shot
   129	*   "telemetry disabled" notice), never POST.
   130	* - empty buffers: short-circuit without a request.
   131	*/
   132	const runFlush = async (args) => {
   133		const { resolved: r, spans, logs, metrics, tracesState, logsState, metricsState, transport, logPrefix, onNoOp } = args;
   134		if (r.noOp) {
   135			spans.drain();
   136			logs.drain();
   137			metrics.drain();
   138			onNoOp();
   139			return;
   140		}
   141		await Promise.all([
   142			flushSignal({
   143				url: r.tracesUrl,
   144				headers: r.headers,
   145				buffer: spans,
   146				body: (items) => makeTracesBody(items, r),
   147				state: tracesState,
   148				signal: "traces",
   149				transport,
   150				logPrefix
   151			}),
   152			flushSignal({
   153				url: r.logsUrl,
   154				headers: r.headers,
   155				buffer: logs,
   156				body: (items) => makeLogsBody(items, r),
   157				state: logsState,
   158				signal: "logs",
   159				transport,
   160				logPrefix
   161			}),
   162			flushSignal({
   163				url: r.metricsUrl,
   164				headers: r.headers,
   165				buffer: metrics,
   166				body: (items) => makeMetricsBody(items, r),
   167				state: metricsState,
   168				signal: "metrics",
   169				transport,
   170				logPrefix
   171			})
   172		]);
   173	};

@@@@@@@@@@ END FILE: sdk-excerpts/01-flush-core-cooldown-serialization.txt

@@@@@@@@@@ FILE: sdk-excerpts/02-buffers-cap-restore.txt (6299 bytes)
Buffers: log/metric/span buffers, 10k cap, restore()
====================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/version-B225m7RW.mjs  lines 293-321  (src/shared/flushable-logger.ts: makeLogBuffer) ---
   293	//#region src/shared/flushable-logger.ts
   294	const MAX_BUFFER$1 = 1e4;
   295	const makeLogBuffer = (options = {}) => {
   296		let buffer = [];
   297		let disabled = false;
   298		const excludeLogSpans = options.excludeLogSpans ?? false;
   299		const logger = Logger.make((logOptions) => {
   300			if (disabled) return;
   301			if (buffer.length >= MAX_BUFFER$1) return;
   302			buffer.push(makeLogRecord(logOptions, excludeLogSpans));
   303		});
   304		return {
   305			loggerLayer: Logger.layer([logger], { mergeWithExisting: true }),
   306			drain: () => {
   307				const items = buffer;
   308				buffer = [];
   309				return items;
   310			},
   311			restore: (items) => {
   312				if (disabled || items.length === 0) return;
   313				buffer = [...items, ...buffer].slice(0, MAX_BUFFER$1);
   314			},
   315			setDisabled: (value) => {
   316				disabled = value;
   317				if (value) buffer = [];
   318			},
   319			size: () => buffer.length
   320		};
   321	};

--- @maple-dev/effect-sdk/dist/version-B225m7RW.mjs  lines 367-419  (src/shared/flushable-metrics.ts: makeMetricBuffer) ---
   367	//#region src/shared/flushable-metrics.ts
   368	/**
   369	* Isolated cumulative registry. Drains only after changes, and `restore`
   370	* re-marks failed exports as pending.
   371	*/
   372	const makeMetricBuffer = () => {
   373		let disabled = false;
   374		let captured = false;
   375		let dirty = false;
   376		/**
   377		* Revoking after capture disables metrics for the page lifetime. Cumulative
   378		* state cannot forget prior samples, and clearing the registry breaks cached
   379		* Metric hooks. A late first grant still works because `captured` is false.
   380		*/
   381		let revoked = false;
   382		const registry = /* @__PURE__ */ new Map();
   383		const set = registry.set.bind(registry);
   384		registry.set = (key, metadata) => {
   385			const hooks = metadata.hooks;
   386			metadata.hooks = {
   387				get: hooks.get,
   388				update: (input, context) => {
   389					if (disabled) return;
   390					captured = true;
   391					hooks.update(input, context);
   392					dirty = true;
   393				},
   394				modify: (input, context) => {
   395					if (disabled) return;
   396					captured = true;
   397					hooks.modify(input, context);
   398					dirty = true;
   399				}
   400			};
   401			return set(key, metadata);
   402		};
   403		const snapshotContext = Context.make(Metric.MetricRegistry, registry);
   404		return {
   405			layer: Layer.succeed(Metric.MetricRegistry, registry),
   406			drain: () => {
   407				if (disabled || revoked || !dirty) return [];
   408				dirty = false;
   409				return [...Metric.snapshotUnsafe(snapshotContext)];
   410			},
   411			restore: (items) => {
   412				if (items.length > 0) dirty = true;
   413			},
   414			setDisabled: (value) => {
   415				if (value && captured) revoked = true;
   416				disabled = value;
   417			}
   418		};
   419	};

--- @maple-dev/effect-sdk/dist/version-B225m7RW.mjs  lines 421-493  (src/shared/flushable-tracer.ts: makeSpanBuffer) ---
   421	//#region src/shared/flushable-tracer.ts
   422	const MAX_BUFFER = 1e4;
   423	const isIgnoredFailure = (error) => Predicate.hasProperty(error, ErrorReporter.ignore) && error[ErrorReporter.ignore] === true;
   424	const isIgnoredSpan = (span) => {
   425		const status = span.status;
   426		if (status._tag !== "Ended") return false;
   427		const exit = status.exit;
   428		if (exit._tag !== "Failure") return false;
   429		if (exit.cause.reasons.some(Cause.isDieReason)) return false;
   430		const failures = exit.cause.reasons.filter(Cause.isFailReason);
   431		return failures.length > 0 && failures.every((reason) => isIgnoredFailure(reason.error));
   432	};
   433	const makeSpanBuffer = (options = {}) => {
   434		let buffer = [];
   435		let disabled = false;
   436		const dropSpan = options.dropSpan;
   437		const anticipatedErrorIdentifiers = options.anticipatedErrorIdentifiers ?? options.anticipatedErrorTags;
   438		const exportFn = (span) => {
   439			if (disabled) return;
   440			if (!span.sampled) return;
   441			if (dropSpan !== void 0 && dropSpan(span.name)) return;
   442			if (isIgnoredSpan(span)) return;
   443			if (buffer.length >= MAX_BUFFER) return;
   444			buffer.push(makeOtlpSpan(span, anticipatedErrorIdentifiers));
   445		};
   446		const tracer = Tracer.make({ span(spanOptions) {
   447			return makeSpan({
   448				...spanOptions,
   449				status: {
   450					_tag: "Started",
   451					startTime: spanOptions.startTime
   452				},
   453				attributes: /* @__PURE__ */ new Map(),
   454				export: exportFn
   455			});
   456		} });
   457		const captureException = (error, captureOptions = {}) => {
   458			if (disabled) return;
   459			const now = BigInt(Date.now()) * 1000000n;
   460			makeSpan({
   461				name: captureOptions.name ?? "exception",
   462				parent: Option.none(),
   463				annotations: Context.empty(),
   464				status: {
   465					_tag: "Started",
   466					startTime: now
   467				},
   468				attributes: new Map(Object.entries(captureOptions.attributes ?? {})),
   469				links: [],
   470				sampled: true,
   471				kind: "internal",
   472				export: exportFn
   473			}).end(now, Exit.failCause(Cause.die(error)));
   474		};
   475		return {
   476			tracerLayer: Layer.succeed(Tracer.Tracer, tracer),
   477			captureException,
   478			drain: () => {
   479				const items = buffer;
   480				buffer = [];
   481				return items;
   482			},
   483			restore: (items) => {
   484				if (disabled || items.length === 0) return;
   485				buffer = [...items, ...buffer].slice(0, MAX_BUFFER);
   486			},
   487			setDisabled: (value) => {
   488				disabled = value;
   489				if (value) buffer = [];
   490			},
   491			size: () => buffer.length
   492		};
   493	};

@@@@@@@@@@ END FILE: sdk-excerpts/02-buffers-cap-restore.txt

@@@@@@@@@@ FILE: sdk-excerpts/03-mapleflush-make-client.txt (14768 bytes)
MapleFlush.make (client): keepalive transport, auto-flush timer, pagehide/visibilitychange flush, dispose
=========================================================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 1984-2175  (src/client/flushable.ts) ---
  1984	//#region src/client/flushable.ts
  1985	var flushable_exports = /* @__PURE__ */ __exportAll({ make: () => make });
  1986	/** Default auto-flush cadence (ms), matching `Otlp.layerJson`'s 5s export interval. */
  1987	const DEFAULT_AUTO_FLUSH_MS = 5e3;
  1988	const browserInstanceId = globalThis.crypto?.randomUUID?.() ?? `browser-${Date.now().toString(36)}-${Math.random().toString(36).slice(2)}`;
  1989	/** `fetch(keepalive)` transport — see file header for why not `sendBeacon`. */
  1990	const keepaliveTransport = { post: async (url, headers, body) => {
  1991		const res = await fetch(url, {
  1992			method: "POST",
  1993			headers,
  1994			body: JSON.stringify(body),
  1995			keepalive: true
  1996		});
  1997		if (!res.ok) throw new Error(`OTLP ${res.status} ${res.statusText}`);
  1998	} };
  1999	const buildBrowserAttributes = (config) => {
  2000		const attributes = {
  2001			"maple.sdk.type": "client",
  2002			"service.instance.id": browserInstanceId
  2003		};
  2004		const nav = browserNavigator();
  2005		if (nav) {
  2006			if (nav.userAgent) attributes["user_agent.original"] = nav.userAgent;
  2007			if (nav.language) attributes["browser.language"] = nav.language;
  2008		}
  2009		if (typeof Intl !== "undefined") {
  2010			const timezone = trySyncOrUndefined(() => Intl.DateTimeFormat().resolvedOptions().timeZone);
  2011			if (timezone) attributes["browser.timezone"] = timezone;
  2012		}
  2013		if (config.environment) {
  2014			attributes["deployment.environment"] = config.environment;
  2015			attributes["deployment.environment.name"] = config.environment;
  2016		}
  2017		if (config.serviceVersion && /^[0-9a-f]{7,40}$/i.test(config.serviceVersion)) attributes["vcs.ref.head.revision"] = config.serviceVersion;
  2018		if (config.serviceNamespace) attributes["service.namespace"] = config.serviceNamespace;
  2019		if (config.attributes) Object.assign(attributes, config.attributes);
  2020		return attributes;
  2021	};
  2022	const make = (config) => {
  2023		const dropPrefixes = config.dropSpanNames;
  2024		const dropSpan = dropPrefixes !== void 0 && dropPrefixes.length > 0 ? (name) => dropPrefixes.some((prefix) => name.startsWith(prefix)) : void 0;
  2025		const anticipatedErrorIdentifiers = [...config.anticipatedErrorIdentifiers ?? [], ...config.anticipatedErrorTags ?? []];
  2026		const anticipatedIdentifiers = anticipatedErrorIdentifiers.length > 0 ? new Set(anticipatedErrorIdentifiers) : void 0;
  2027		const clientSession = startClientSession({
  2028			endpoint: config.endpoint,
  2029			ingestKey: config.ingestKey,
  2030			serviceName: config.serviceName,
  2031			environment: config.environment,
  2032			serviceVersion: config.serviceVersion,
  2033			replay: config.replay,
  2034			emitSessionMeta: config.emitSessionMeta,
  2035			privacy: config.privacy
  2036		});
  2037		const spans = makeSpanBuffer({
  2038			dropSpan,
  2039			anticipatedErrorIdentifiers: anticipatedIdentifiers
  2040		});
  2041		const logs = makeLogBuffer({ excludeLogSpans: config.excludeLogSpans });
  2042		const metrics = makeMetricBuffer();
  2043		const setCaptureAllowed = (allowed) => {
  2044			spans.setDisabled(!allowed);
  2045			logs.setDisabled(!allowed);
  2046			metrics.setDisabled(!allowed);
  2047		};
  2048		setCaptureAllowed(hasConsent());
  2049		const stopConsentListener = config.privacy?.requireConsent ? onConsentChange(setCaptureAllowed) : () => {};
  2050		const layer = withSessionLink(Layer.mergeAll(spans.tracerLayer, logs.loggerLayer, metrics.layer));
  2051		const resource = {
  2052			endpoint: config.endpoint,
  2053			ingestKey: config.ingestKey ? Redacted.make(config.ingestKey) : void 0,
  2054			resource: {
  2055				serviceName: config.serviceName,
  2056				serviceVersion: config.serviceVersion,
  2057				attributes: buildBrowserAttributes(config)
  2058			}
  2059		};
  2060		const resolved = buildResolved(resource, {
  2061			tracesPath: config.tracesPath,
  2062			logsPath: config.logsPath,
  2063			metricsPath: config.metricsPath,
  2064			userAgent: `maple-effect-sdk-client/${SDK_VERSION}`
  2065		});
  2066		const tracesState = { disabledUntil: 0 };
  2067		const logsState = { disabledUntil: 0 };
  2068		const metricsState = { disabledUntil: 0 };
  2069		const noOpNotice = makeNoOpNotice("[MapleClientSDK]", "pass `ingestKey` to enable");
  2070		const flush = makeSerializedFlush(guardFlush("[MapleClientSDK]", async () => {
  2071			if (!hasConsent()) {
  2072				spans.drain();
  2073				logs.drain();
  2074				metrics.drain();
  2075				return;
  2076			}
  2077			await runFlush({
  2078				resolved,
  2079				spans,
  2080				logs,
  2081				metrics,
  2082				tracesState,
  2083				logsState,
  2084				metricsState,
  2085				transport: keepaliveTransport,
  2086				logPrefix: "[MapleClientSDK]",
  2087				onNoOp: noOpNotice
  2088			});
  2089		}));
  2090		const intervalMs = config.autoFlushInterval === void 0 ? DEFAULT_AUTO_FLUSH_MS : config.autoFlushInterval === false ? 0 : config.autoFlushInterval;
  2091		let timer;
  2092		if (intervalMs > 0) {
  2093			timer = setInterval(() => {
  2094				flush();
  2095			}, intervalMs);
  2096			timer.unref?.();
  2097		}
  2098		/**
  2099		* One error reaching two paths must still be one issue. React rethrows a
  2100		* boundary-caught error in development, so a boundary that reports it *and*
  2101		* `window.onerror` would otherwise fingerprint the same crash twice.
  2102		*/
  2103		const reported = /* @__PURE__ */ new WeakSet();
  2104		const captureException = (error, options = {}) => {
  2105			if (typeof error === "object" && error !== null) {
  2106				if (reported.has(error)) return;
  2107				reported.add(error);
  2108			}
  2109			const page = globalThis.location?.href;
  2110			spans.captureException(error, {
  2111				...options,
  2112				attributes: {
  2113					...page !== void 0 ? { "url.full": page } : void 0,
  2114					...options.attributes
  2115				}
  2116			});
  2117		};
  2118		const onWindowError = (event) => {
  2119			const error = event.error ?? (event.message && event.filename ? new Error(event.message) : void 0);
  2120			if (error === void 0) return;
  2121			captureException(error, {
  2122				name: "browser.uncaught_error",
  2123				attributes: {
  2124					"maple.exception.source": "window.onerror",
  2125					...event.filename ? { "code.file.path": event.filename } : void 0,
  2126					...event.lineno ? { "code.line.number": event.lineno } : void 0
  2127				}
  2128			});
  2129		};
  2130		const onUnhandledRejection = (event) => {
  2131			captureException(event.reason, {
  2132				name: "browser.unhandled_rejection",
  2133				attributes: { "maple.exception.source": "unhandledrejection" }
  2134			});
  2135		};
  2136		const canCaptureGlobals = (config.captureGlobalErrors ?? true) && typeof globalThis.addEventListener === "function";
  2137		if (canCaptureGlobals) {
  2138			globalThis.addEventListener("error", onWindowError);
  2139			globalThis.addEventListener("unhandledrejection", onUnhandledRejection);
  2140		}
  2141		const onPageHide = () => {
  2142			flush();
  2143		};
  2144		const onVisibilityChange = () => {
  2145			if (browserDocument()?.visibilityState === "hidden") flush();
  2146		};
  2147		const canListen = (config.flushOnUnload ?? true) && typeof globalThis.addEventListener === "function";
  2148		if (canListen) {
  2149			globalThis.addEventListener("pagehide", onPageHide);
  2150			globalThis.addEventListener("visibilitychange", onVisibilityChange);
  2151		}
  2152		const dispose = async () => {
  2153			if (timer !== void 0) {
  2154				clearInterval(timer);
  2155				timer = void 0;
  2156			}
  2157			if (canListen) {
  2158				globalThis.removeEventListener("pagehide", onPageHide);
  2159				globalThis.removeEventListener("visibilitychange", onVisibilityChange);
  2160			}
  2161			if (canCaptureGlobals) {
  2162				globalThis.removeEventListener("error", onWindowError);
  2163				globalThis.removeEventListener("unhandledrejection", onUnhandledRejection);
  2164			}
  2165			stopConsentListener();
  2166			await flush();
  2167			await clientSession.stop();
  2168		};
  2169		return {
  2170			layer,
  2171			captureException,
  2172			flush,
  2173			dispose
  2174		};
  2175	};

--- @maple-dev/effect-sdk/dist/index-CM1baBqK.d.mts  lines 123-225  (type declarations: MapleClientFlushableConfig / FlushableTelemetry) ---
   123	interface MapleClientFlushableConfig {
   124	  /** Service name reported in traces, logs, and metrics. */
   125	  readonly serviceName: string;
   126	  /** Maple ingest endpoint URL. */
   127	  readonly endpoint: string;
   128	  /** Maple ingest key. When unset, the preset runs in no-op mode. */
   129	  readonly ingestKey?: string | undefined;
   130	  /** Service version or commit SHA. */
   131	  readonly serviceVersion?: string | undefined;
   132	  /**
   133	   * Logical group this service belongs to, emitted as the OTel
   134	   * `service.namespace` resource attribute. Optional — only stamped when set.
   135	   */
   136	  readonly serviceNamespace?: string | undefined;
   137	  /** Deployment environment (e.g. "production", "staging"). */
   138	  readonly environment?: string | undefined;
   139	  /** Additional resource attributes (highest precedence). */
   140	  readonly attributes?: Record<string, unknown> | undefined;
   141	  /** Skip Effect log spans in OTLP log attributes. Default `false`. */
   142	  readonly excludeLogSpans?: boolean | undefined;
   143	  /** Span name prefixes to drop before OTLP export. */
   144	  readonly dropSpanNames?: ReadonlyArray<string> | undefined;
   145	  /**
   146	   * Stable `_tag` / `Error.name` identifiers of anticipated 4xx failures. Spans
   147	   * failing entirely with these export as status `Ok` (no `exception` event),
   148	   * so they stay visible but never count as errors.
   149	   */
   150	  readonly anticipatedErrorIdentifiers?: ReadonlyArray<string> | undefined;
   151	  /** @deprecated Use `anticipatedErrorIdentifiers`. */
   152	  readonly anticipatedErrorTags?: ReadonlyArray<string> | undefined;
   153	  /** OTLP traces path appended to `endpoint`. Default `/v1/traces`. */
   154	  readonly tracesPath?: string | undefined;
   155	  /** OTLP logs path appended to `endpoint`. Default `/v1/logs`. */
   156	  readonly logsPath?: string | undefined;
   157	  /** OTLP metrics path appended to `endpoint`. Default `/v1/metrics`. */
   158	  readonly metricsPath?: string | undefined;
   159	  /**
   160	   * Background auto-flush cadence in milliseconds. Default `5000`. Set to `0`
   161	   * or `false` to disable and flush purely on demand (note: the in-memory
   162	   * buffer caps at 10k items, so a long-lived tab that never flushes will
   163	   * eventually drop new spans).
   164	   */
   165	  readonly autoFlushInterval?: number | false | undefined;
   166	  /**
   167	   * Register `pagehide` + `visibilitychange→hidden` listeners that flush the
   168	   * buffer before the tab goes away. Default `true`. No-ops when there's no
   169	   * `addEventListener` (SSR / non-DOM runtime).
   170	   */
   171	  readonly flushOnUnload?: boolean | undefined;
   172	  /**
   173	   * Post session metadata rows for the standalone session so it appears in
   174	   * Maple's Sessions UI (list entry + linked traces, no replay recording).
   175	   * Default `true`; no-ops when `@maple-dev/browser` is on the page (it owns
   176	   * the session rows), without a browser DOM, or without an ingest key.
   177	   */
   178	  readonly emitSessionMeta?: boolean | undefined;
   179	  /**
   180	   * rrweb session replay for this app, recorded by the shared Maple replay
   181	   * engine (loaded lazily in a code-split chunk). Default enabled with
   182	   * sampleRate 1 and inputs masked — set `{ enabled: false }` to opt out.
   183	   */
   184	  readonly replay?: ClientReplayConfig | undefined;
   185	  /**
   186	   * Capture uncaught errors and unhandled promise rejections from the page and
   187	   * record them as error spans. Default `true`.
   188	   *
   189	   * Without this the SDK only ever sees failures that happened *inside* an
   190	   * Effect span, which in a browser is the minority of them — a React render
   191	   * crash, a throw in an event handler and a floating rejected promise all
   192	   * bypass Effect entirely and would otherwise never reach Maple.
   193	   */
   194	  readonly captureGlobalErrors?: boolean | undefined;
   195	  /**
   196	   * Consent gating, persistent-visitor-id storage, and whether `identify()`'s
   197	   * email reaches the warehouse. Defaults capture everything except where a
   198	   * browser signals otherwise: Global Privacy Control suppresses the
   199	   * persistent visitor id (capture itself continues, anonymously).
   200	   */
   201	  readonly privacy?: PrivacyOptions | undefined;
   202	}
   203	interface FlushableTelemetry {
   204	  /**
   205	   * Effect Layer installing the buffer-backed OTLP tracer (with replay-session
   206	   * linking) + Effect logger. Must live in the same runtime as your
   207	   * instrumented code.
   208	   */
   209	  readonly layer: Layer.Layer<never>;
   210	  /**
   211	   * Record an error that never passed through an Effect span — the escape
   212	   * hatch for the places a browser throws outside Effect. The canonical caller
   213	   * is a React error boundary, which catches the error and, unless it reports
   214	   * it here, is the reason nobody ever hears about the crash.
   215	   *
   216	   * BOUNDARY: a thrown value is unparsed by definition — JavaScript can throw
   217	   * anything. It is narrowed on the way into the exception event.
   218	   */
   219	  readonly captureException: (error: unknown, options?: CaptureExceptionOptions) => void;
   220	  /** Drain the buffers and POST them now (keepalive). Never rejects. */
   221	  readonly flush: () => Promise<void>;
   222	  /** Remove unload listeners, stop the auto-flush timer, then do one final flush. */
   223	  readonly dispose: () => Promise<void>;
   224	}
   225	declare const make: (config: MapleClientFlushableConfig) => FlushableTelemetry;

@@@@@@@@@@ END FILE: sdk-excerpts/03-mapleflush-make-client.txt

@@@@@@@@@@ FILE: sdk-excerpts/04-replay-transport-keepalive-budget.txt (6957 bytes)
Session/replay transport: MAX_KEEPALIVE_BYTES, keepaliveFor(), warnDropped(), postSessionMeta/Events/Blob
=========================================================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 569-687  (../browser-session/src/platform/transport.ts) ---
   569	//#region ../browser-session/src/platform/transport.ts
   570	/**
   571	* Header stamped on every request to ingest: `<sdk-name>/<version>`, e.g.
   572	* `maple-browser/0.3.0`. Browsers refuse to let a page set `user-agent`, so
   573	* without this a rejected request carried nothing that said which SDK build
   574	* produced it. Ingest records it as `maple.sdk` on the request span. Every
   575	* gateway that receives browser traffic must allow it in CORS (`apps/ingest`
   576	* and the CLI's local listener do) — a header the SDK always sends and
   577	* preflight refuses blocks the whole SDK, not just the header.
   578	*/
   579	const SDK_HINT_HEADER = "x-maple-sdk";
   580	/** Auth + identity headers shared by every ingest request. */
   581	function ingestHeaders(config) {
   582		return {
   583			Authorization: `Bearer ${config.ingestKey}`,
   584			[SDK_HINT_HEADER]: config.sdk
   585		};
   586	}
   587	const lastWarnAt = /* @__PURE__ */ new Map();
   588	function warnDropped(what, error) {
   589		const now = Date.now();
   590		if (now - (lastWarnAt.get(what) ?? 0) < 3e4) return;
   591		lastWarnAt.set(what, now);
   592		console.warn(`[maple] session replay ${what} failed (dropping; will retry on next chunk):`, error);
   593	}
   594	/**
   595	* Largest body we will hand to a `keepalive` fetch.
   596	*
   597	* The Fetch spec caps the *combined* inflight keepalive body at 64 KiB, and on
   598	* the way out we issue up to three of these at once (metadata row, final events
   599	* batch, last replay chunk). Over the budget the browser rejects the request
   600	* outright, so a normal request — which the page may or may not survive long
   601	* enough to finish — is strictly the better bet than a guaranteed rejection.
   602	*/
   603	const MAX_KEEPALIVE_BYTES = 49152;
   604	/** Whether a body of `bytes` may still ride the keepalive budget. */
   605	function keepaliveFor(requested, bytes) {
   606		return requested && bytes <= MAX_KEEPALIVE_BYTES;
   607	}
   608	/**
   609	* Smallest well-formed gzip stream: a 10-byte header plus an 8-byte trailer.
   610	* Anything shorter cannot carry a complete member, whatever it contains.
   611	*/
   612	const GZIP_MIN_BYTES = 18;
   613	/**
   614	* gzip a byte buffer using the native CompressionStream (no library).
   615	*
   616	* Throws rather than returning a partial stream. The write and close promises
   617	* used to be discarded, so a page torn down mid-compression still resolved
   618	* `arrayBuffer()` — with only the bytes already flushed — and the caller posted
   619	* a 3-byte header stub as if it were a complete chunk. Ingest rejected those as
   620	* corrupt gzip; they were 82% of all replay-blob failures, overwhelmingly from
   621	* crawlers, which tear the page down on nearly every session.
   622	*/
   623	async function gzip(bytes) {
   624		const stream = new CompressionStream("gzip");
   625		const writer = stream.writable.getWriter();
   626		const written = writer.write(bytes).then(() => writer.close());
   627		const [buffer] = await Promise.all([new Response(stream.readable).arrayBuffer(), written]);
   628		const out = new Uint8Array(buffer);
   629		if (out.length < GZIP_MIN_BYTES || out[0] !== 31 || out[1] !== 139 || out[2] !== 8) throw new Error(`gzip produced ${out.length} bytes, not a complete gzip stream`);
   630		return out;
   631	}
   632	/** POST session metadata (NDJSON, single row). `keepalive` for the final unload write. */
   633	async function postSessionMeta(config, row, keepalive = false) {
   634		const body = `${JSON.stringify(row)}\n`;
   635		await fetch(`${config.endpoint}/v1/sessionReplays/meta`, {
   636			method: "POST",
   637			headers: {
   638				...ingestHeaders(config),
   639				"content-type": "application/x-ndjson"
   640			},
   641			body,
   642			keepalive: keepaliveFor(keepalive, body.length)
   643		}).catch((error) => {
   644			warnDropped("metadata POST", error);
   645		});
   646	}
   647	/** POST distilled session events (NDJSON, one row per event). Best-effort. */
   648	async function postSessionEvents(config, rows, keepalive = false) {
   649		if (rows.length === 0) return;
   650		const body = `${rows.map((row) => JSON.stringify(row)).join("\n")}\n`;
   651		await fetch(`${config.endpoint}/v1/sessionEvents`, {
   652			method: "POST",
   653			headers: {
   654				...ingestHeaders(config),
   655				"content-type": "application/x-ndjson"
   656			},
   657			body,
   658			keepalive: keepaliveFor(keepalive, body.length)
   659		}).catch((error) => {
   660			warnDropped("events POST", error);
   661		});
   662	}
   663	const SESSION_EXHAUSTED_STATUS = 413;
   664	/** POST a gzipped rrweb event chunk. */
   665	async function postSessionBlob(config, meta, gzipped, keepalive = false) {
   666		try {
   667			const response = await fetch(`${config.endpoint}/v1/sessionReplays/blob`, {
   668				method: "POST",
   669				headers: {
   670					...ingestHeaders(config),
   671					"content-type": "application/octet-stream",
   672					"x-maple-session-id": meta.sessionId,
   673					"x-maple-chunk-seq": String(meta.chunkSeq),
   674					"x-maple-is-checkpoint": meta.isCheckpoint ? "1" : "0",
   675					"x-maple-event-count": String(meta.eventCount),
   676					"x-maple-duration-ms": String(meta.durationMs)
   677				},
   678				body: gzipped,
   679				keepalive: keepaliveFor(keepalive, gzipped.byteLength)
   680			});
   681			if (response.ok) return "accepted";
   682			return response.status === SESSION_EXHAUSTED_STATUS ? "exhausted" : "rejected";
   683		} catch (error) {
   684			warnDropped("blob PUT", error);
   685			return "failed";
   686		}
   687	}

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 796-808  (../browser-session/src/events/meta-row.ts: postSessionMetaRow (metadata-only sessions)) ---
   796	/** POST one session metadata row (NDJSON). Best-effort — never throws. */
   797	async function postSessionMetaRow(target, row, keepalive = false) {
   798		const body = `${JSON.stringify(row)}\n`;
   799		await fetch(`${target.endpoint.replace(/\/$/, "")}/v1/sessionReplays/meta`, {
   800			method: "POST",
   801			headers: {
   802				...ingestHeaders(target),
   803				"content-type": "application/x-ndjson"
   804			},
   805			body,
   806			keepalive: keepaliveFor(keepalive, body.length)
   807		}).catch(() => {});
   808	}

@@@@@@@@@@ END FILE: sdk-excerpts/04-replay-transport-keepalive-budget.txt

@@@@@@@@@@ FILE: sdk-excerpts/05-session-event-sink.txt (4230 bytes)
Session event sink: 5 s non-keepalive flush, drop-on-failure
============================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 1123-1220  (../browser-session/src/events/events-sink.ts) ---
  1123	//#region ../browser-session/src/events/events-sink.ts
  1124	const FLUSH_INTERVAL_MS = 5e3;
  1125	const FLUSH_BYTES = 65536;
  1126	/**
  1127	* The sink is a per-session singleton, published on `globalThis` rather than a
  1128	* module-level variable.
  1129	*
  1130	* Two consumers can end up with separate copies of this module in one page (an
  1131	* app bundling both `@maple-dev/browser` and the Effect SDK, or a
  1132	* lazily-imported replay chunk). Two sinks would mean two `seq` counters
  1133	* starting at 0, and `Seq` is part of the `session_events` sorting key — the
  1134	* rows would collide. One global owner avoids that.
  1135	*/
  1136	const SINK_KEY = "__MAPLE_SESSION_EVENT_SINK__";
  1137	function holder() {
  1138		return globalThis;
  1139	}
  1140	/** The live sink, if one has been started. */
  1141	function getActiveSink() {
  1142		return holder()[SINK_KEY]?.sink;
  1143	}
  1144	/**
  1145	* Start (or reuse) the distilled-event sink for a session.
  1146	*
  1147	* Runs on **every** page load, not just sampled-for-replay ones: page views and
  1148	* `track()` calls are the analytics substrate, and gating them behind replay
  1149	* sampling would make unique visitors and top pages a sample rather than a
  1150	* count. The rrweb recorder stays sampled — it is the expensive part.
  1151	*/
  1152	function startEventSink(config, sessionId) {
  1153		const existing = holder()[SINK_KEY];
  1154		if (existing && existing.sessionId === sessionId) return existing.sink;
  1155		if (existing) {
  1156			existing.sink.flush();
  1157			existing.sink.stop();
  1158		}
  1159		let buffer = [];
  1160		let bufferBytes = 0;
  1161		let seq = 0;
  1162		let pageViews = 0;
  1163		let clickCount = 0;
  1164		let errorCount = 0;
  1165		const flush = async (keepalive = false) => {
  1166			if (buffer.length === 0) return;
  1167			const batch = buffer;
  1168			buffer = [];
  1169			bufferBytes = 0;
  1170			await postSessionEvents(config, batch.map(({ ev, seq }) => toRow(config, sessionId, ev, seq)), keepalive);
  1171		};
  1172		const emit = (ev) => {
  1173			const session = markActivity();
  1174			if (session && session.id !== sessionId) {
  1175				const nextSink = startEventSink(config, session.id);
  1176				if (ev.type !== "navigation") nextSink.emit(ev);
  1177				return;
  1178			}
  1179			if (ev.type === "navigation") {
  1180				pageViews++;
  1181				noteNavigation(ev.url ?? (typeof location !== "undefined" ? location.href : ""));
  1182			} else if (ev.type === "click") clickCount++;
  1183			else if (ev.type === "error" || ev.type === "console" && ev.level === "error") errorCount++;
  1184			buffer.push({
  1185				ev,
  1186				seq: seq++
  1187			});
  1188			bufferBytes += approximateSize(ev);
  1189			if (bufferBytes >= FLUSH_BYTES) flush();
  1190		};
  1191		const stopNavigation = installNavigationObserver((url) => {
  1192			emit({
  1193				type: "navigation",
  1194				url
  1195			});
  1196		});
  1197		const stopBaselineCapture = startBaselineCapture(emit, config.maskAllText);
  1198		const flushTimer = setInterval(() => void flush(), FLUSH_INTERVAL_MS);
  1199		const sink = {
  1200			sessionId,
  1201			emit,
  1202			flush,
  1203			stop: () => {
  1204				clearInterval(flushTimer);
  1205				stopNavigation();
  1206				stopBaselineCapture();
  1207				if (holder()[SINK_KEY]?.sink === sink) holder()[SINK_KEY] = void 0;
  1208			},
  1209			getPageViews: () => pageViews,
  1210			getClickCount: () => clickCount,
  1211			getErrorCount: () => errorCount,
  1212			ignoreUrl: (url) => url.startsWith(`${config.endpoint}/v1/`)
  1213		};
  1214		holder()[SINK_KEY] = {
  1215			sessionId,
  1216			sink
  1217		};
  1218		drainPending(sink);
  1219		return sink;
  1220	}

@@@@@@@@@@ END FILE: sdk-excerpts/05-session-event-sink.txt

@@@@@@@@@@ FILE: sdk-excerpts/06-session-lifecycle-unload.txt (10642 bytes)
Session lifecycle: pagehide/visibilitychange 'ended' row (keepalive) + metadata-only session + trace-id cap comment
===================================================================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 1410-1611  (../browser-session/src/session/lifecycle.ts + metadata-session.ts) ---
  1410	//#region ../browser-session/src/session/lifecycle.ts
  1411	/**
  1412	* How often a visible tab re-posts its `active` row.
  1413	*
  1414	* This is what makes exit page, page views and duration survive a tab killed
  1415	* without an unload beacon — otherwise the only surviving row is the v1 row
  1416	* posted at session start, whose counters are all zero and therefore read as a
  1417	* bounce. 60s is the floor worth using: the table is a ReplacingMergeTree, so
  1418	* every heartbeat is an unmerged part until the next merge. Billing meters only
  1419	* `version == 1` rows, so heartbeats do not double-bill.
  1420	*/
  1421	const HEARTBEAT_INTERVAL_MS = 6e4;
  1422	/**
  1423	* Drive one session's metadata lifecycle. Returns undefined outside a browser;
  1424	* sampling and consent are the caller's decisions.
  1425	*/
  1426	function startSessionLifecycle(options, hooks) {
  1427		if (typeof window === "undefined") return void 0;
  1428		let current = getSession();
  1429		let stopped = false;
  1430		/** Whether a run is live: capture started, `active` posted, heartbeat armed. */
  1431		let running = false;
  1432		let heartbeat;
  1433		let clickCountBase = 0;
  1434		let errorCountBase = 0;
  1435		let sinkClickCountAtStart = 0;
  1436		let sinkErrorCountAtStart = 0;
  1437		/**
  1438		* The persisted record of the session this lifecycle owns.
  1439		*
  1440		* `current` fixes only the identity of the run — counters, `lastUrl` and
  1441		* `lastActivityAt` move underneath it as the page is used, so a row built
  1442		* from the captured object alone reports the exit path and page views of the
  1443		* session's *first* page forever. Reading storage back per row keeps them
  1444		* live, and the id check keeps a rotation from attributing the incoming
  1445		* session's context to the outgoing session's `ended` row.
  1446		*/
  1447		const liveRecord = () => {
  1448			const stored = peekSession();
  1449			return stored?.id === current.id ? stored : current;
  1450		};
  1451		/**
  1452		* Re-baseline the cumulative counters against the persisted record.
  1453		*
  1454		* The record is what carries counts across a reload and across a hide/resume
  1455		* cycle (each `ended` row writes them back via `noteCounts`), while live
  1456		* capture restarts from zero every run — so a run's row is the persisted base
  1457		* plus what capture has seen since this run began.
  1458		*/
  1459		const rebaseCounts = (record) => {
  1460			clickCountBase = record.clickCount ?? 0;
  1461			errorCountBase = record.errorCount ?? 0;
  1462			const sink = getActiveSink();
  1463			sinkClickCountAtStart = sink?.getClickCount() ?? 0;
  1464			sinkErrorCountAtStart = sink?.getErrorCount() ?? 0;
  1465		};
  1466		const countsFor = (record) => {
  1467			const sink = getActiveSink();
  1468			const sinkMatches = sink !== void 0 && sink.sessionId === record.id;
  1469			const clicks = hooks.clicksSinceStart?.() ?? (sinkMatches ? sink.getClickCount() - sinkClickCountAtStart : 0);
  1470			const errors = sinkMatches ? sink.getErrorCount() - sinkErrorCountAtStart : 0;
  1471			const counts = {
  1472				clickCount: clickCountBase + Math.max(0, clicks),
  1473				errorCount: errorCountBase + Math.max(0, errors),
  1474				pageViews: record.pageViews ?? (sinkMatches ? sink.getPageViews() : 0)
  1475			};
  1476			if (peekSession()?.id === record.id) noteCounts({
  1477				clickCount: counts.clickCount,
  1478				errorCount: counts.errorCount
  1479			});
  1480			return counts;
  1481		};
  1482		const post = (status, keepalive) => {
  1483			const record = liveRecord();
  1484			const counts = countsFor(record);
  1485			hooks.post(buildSessionMetaRow({
  1486				sessionId: record.id,
  1487				startedAt: new Date(record.startedAt),
  1488				version: nextMetaVersion(),
  1489				status,
  1490				serviceName: options.serviceName,
  1491				identity: options.getIdentity?.(),
  1492				captureUserEmail: options.captureUserEmail,
  1493				environment: options.environment,
  1494				serviceVersion: options.serviceVersion,
  1495				visitorId: getVisitorId(),
  1496				visitorIsNew: record.visitorIsNew === true,
  1497				visitorIdPersisted: isVisitorIdPersisted(),
  1498				entry: entryContextOf(record),
  1499				lastUrl: record.lastUrl,
  1500				clickCount: counts.clickCount,
  1501				pageViews: counts.pageViews,
  1502				errorCount: counts.errorCount,
  1503				traceIds: status === "ended" ? options.getTraceIds?.(record.id) : void 0,
  1504				recorded: hooks.recorded
  1505			}), keepalive);
  1506		};
  1507		const stopHeartbeat = () => {
  1508			if (heartbeat === void 0) return;
  1509			clearInterval(heartbeat);
  1510			heartbeat = void 0;
  1511		};
  1512		const startRun = () => {
  1513			if (stopped || running) return;
  1514			running = true;
  1515			const record = liveRecord();
  1516			rebaseCounts(record);
  1517			hooks.onStart?.(record);
  1518			post("active", false);
  1519			heartbeat = setInterval(() => {
  1520				if (typeof document !== "undefined" && document.visibilityState !== "visible") return;
  1521				if (isSessionExpired(liveRecord())) {
  1522					endRun({
  1523						flush: true,
  1524						keepalive: false
  1525					});
  1526					return;
  1527				}
  1528				post("active", false);
  1529			}, HEARTBEAT_INTERVAL_MS);
  1530		};
  1531		const endRun = (suspend) => {
  1532			if (!running) return;
  1533			running = false;
  1534			stopHeartbeat();
  1535			if (suspend.flush) post("ended", suspend.keepalive);
  1536			return hooks.onSuspend?.(suspend);
  1537		};
  1538		const stopRotationListener = onSessionRotate((previous, next) => {
  1539			if (stopped || previous.id !== current.id) return;
  1540			endRun({
  1541				flush: true,
  1542				keepalive: false
  1543			});
  1544			current = next;
  1545			hooks.onSessionChange?.(next.id);
  1546			queueMicrotask(() => startRun());
  1547		});
  1548		const onPageHide = () => void endRun({
  1549			flush: true,
  1550			keepalive: true
  1551		});
  1552		const onVisibilityChange = () => {
  1553			const doc = browserDocument$1();
  1554			if (!doc) return;
  1555			if (doc.visibilityState === "hidden") {
  1556				endRun({
  1557					flush: true,
  1558					keepalive: true
  1559				});
  1560				return;
  1561			}
  1562			if (isSessionExpired(liveRecord())) {
  1563				getSession();
  1564				return;
  1565			}
  1566			startRun();
  1567		};
  1568		const visibilityTarget = typeof document !== "undefined" && typeof document.addEventListener === "function" ? document : typeof globalThis.addEventListener === "function" ? globalThis : void 0;
  1569		const pageHideTarget = typeof globalThis.addEventListener === "function" ? globalThis : void 0;
  1570		startRun();
  1571		visibilityTarget?.addEventListener("visibilitychange", onVisibilityChange);
  1572		pageHideTarget?.addEventListener("pagehide", onPageHide);
  1573		return {
  1574			get sessionId() {
  1575				return current.id;
  1576			},
  1577			shutdown: async (shutdownOptions) => {
  1578				if (stopped) return;
  1579				stopped = true;
  1580				stopRotationListener();
  1581				visibilityTarget?.removeEventListener("visibilitychange", onVisibilityChange);
  1582				pageHideTarget?.removeEventListener("pagehide", onPageHide);
  1583				const flush = shutdownOptions?.flush ?? true;
  1584				await endRun({
  1585					flush,
  1586					keepalive: flush
  1587				});
  1588			}
  1589		};
  1590	}
  1591	//#endregion
  1592	//#region ../browser-session/src/session/metadata-session.ts
  1593	/**
  1594	* Metadata-only session lifecycle used whenever rrweb is disabled or unsampled.
  1595	* It is the shared lifecycle driver with no capture of its own: the distilled
  1596	* event sink runs on every page load regardless, so this owner only has to post
  1597	* rows and flush that sink on the way out.
  1598	*/
  1599	function startMetadataSession(options) {
  1600		return startSessionLifecycle(options, {
  1601			recorded: false,
  1602			post: (row, keepalive) => {
  1603				postSessionMetaRow(options, row, keepalive);
  1604			},
  1605			onSuspend: ({ flush, keepalive }) => {
  1606				if (flush && keepalive) getActiveSink()?.flush(true);
  1607			},
  1608			onSessionChange: options.onSessionChange
  1609		});
  1610	}
  1611	//#endregion

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 1612-1640  (../browser-session/src/session/sink.ts (comment about the shared 64 KiB budget)) ---
  1612	//#region ../browser-session/src/session/sink.ts
  1613	const SESSION_SINK_KEY = "__MAPLE_BROWSER_SESSION__";
  1614	const observedTraceIdsBySession = /* @__PURE__ */ new Map();
  1615	/**
  1616	* Ceiling on trace ids retained per session.
  1617	*
  1618	* A session lives up to 24h and every span feeds this, so it is otherwise
  1619	* unbounded — and the whole set is serialized into the `ended` metadata row,
  1620	* which is written with `keepalive` on the way out. The Fetch spec caps the
  1621	* *combined* keepalive body across in-flight requests at 64 KiB, shared here
  1622	* with the final events flush and the last replay chunk, so an app emitting a
  1623	* span a second would silently lose its entire session row after ~30 minutes.
  1624	*
  1625	* Keep-first rather than keep-last: the ids are a join key for "show me this
  1626	* session's traces", the UI paginates them anyway, and dropping the tail of a
  1627	* long session is a smaller loss than dropping the row.
  1628	*/
  1629	const MAX_TRACE_IDS_PER_SESSION = 200;
  1630	/** Record a trace id seen during the session. Idempotent per id. */
  1631	function recordTraceId(traceId, sessionId = readSessionSink()?.sessionId) {
  1632		if (!sessionId) return;
  1633		let ids = observedTraceIdsBySession.get(sessionId);
  1634		if (!ids) {
  1635			ids = /* @__PURE__ */ new Set();
  1636			observedTraceIdsBySession.set(sessionId, ids);
  1637		}
  1638		if (ids.size >= MAX_TRACE_IDS_PER_SESSION && !ids.has(traceId)) return;
  1639		ids.add(traceId);
  1640	}

@@@@@@@@@@ END FILE: sdk-excerpts/06-session-lifecycle-unload.txt

@@@@@@@@@@ FILE: sdk-excerpts/07-client-session-startup.txt (9123 bytes)
startClientSession: side effects at MapleFlush.make() time, replay dynamic import, stop()
=========================================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 1750-1937  (src/client/standalone-session.ts + replay-loader.ts) ---
  1750	//#endregion
  1751	//#region src/client/standalone-session.ts
  1752	/** Trace ids observed per standalone session — attached to its ended row. */
  1753	const observedBySession = /* @__PURE__ */ new Map();
  1754	let current;
  1755	/**
  1756	* The page-level metadata session, handed out as reference-counted leases.
  1757	*
  1758	* The refcount is load-bearing, not defensive scaffolding: `Maple.layer` starts
  1759	* a client session per *runtime*, and two runtimes can overlap on one page —
  1760	* `ManagedRuntime.dispose()` resolves asynchronously, so a React StrictMode
  1761	* double-mount (or an HMR remount) builds the replacement runtime before the
  1762	* outgoing one's release finalizer runs. Both then hold a lease on the same
  1763	* singleton. Without the count, the outgoing runtime's `shutdown` posts the
  1764	* `ended` row and clears `current` while the live runtime is still holding the
  1765	* handle — the session dies mid-page and nothing restarts it, because
  1766	* `setupStandaloneSession` only runs on runtime start. Two `MapleFlush.make()`
  1767	* calls in one page reproduce the same overlap.
  1768	*
  1769	* Within a single `startClientSession` controller leases never overlap
  1770	* (`stopRuntime` calls the outgoing `shutdown` synchronously, before any await,
  1771	* and `startRuntime` refuses to run while a runtime exists), so the count only
  1772	* ever exceeds 1 across controllers. The per-lease `released` flag keeps a
  1773	* double `shutdown` of one lease from decrementing twice and stranding another
  1774	* holder. Covered by the overlap cases in `standalone-session.test.ts`.
  1775	*/
  1776	const leaseCurrent = () => {
  1777		if (!current) return void 0;
  1778		current.references++;
  1779		const owned = current.handle;
  1780		let released = false;
  1781		return {
  1782			get sessionId() {
  1783				return owned.sessionId;
  1784			},
  1785			shutdown: async (options) => {
  1786				if (released) return;
  1787				released = true;
  1788				if (!current || current.handle !== owned) return;
  1789				current.references--;
  1790				if (current.references > 0) return;
  1791				current = void 0;
  1792				await owned.shutdown(options);
  1793			}
  1794		};
  1795	};
  1796	/** Record a span against the session resolved by the shared span decorator. */
  1797	const noteStandaloneSpan = (sessionId, traceId) => {
  1798		let ids = observedBySession.get(sessionId);
  1799		if (!ids) {
  1800			ids = /* @__PURE__ */ new Set();
  1801			observedBySession.set(sessionId, ids);
  1802		}
  1803		ids.add(traceId);
  1804	};
  1805	/**
  1806	* Start metadata-only emission when the browser SDK is not already the owner.
  1807	* Idempotent per Effect client runtime.
  1808	*/
  1809	const setupStandaloneSession = (options) => {
  1810		if (typeof window === "undefined" || typeof document === "undefined" || !options.ingestKey || readSessionSink()) return void 0;
  1811		if (current) return leaseCurrent();
  1812		const handle = startMetadataSession({
  1813			endpoint: options.endpoint,
  1814			ingestKey: options.ingestKey,
  1815			sdk: CLIENT_SDK_HINT,
  1816			serviceName: options.serviceName,
  1817			environment: options.environment,
  1818			serviceVersion: options.serviceVersion,
  1819			captureUserEmail: options.captureUserEmail,
  1820			getIdentity: getCurrentIdentity,
  1821			getTraceIds: (sessionId) => Array.from(observedBySession.get(sessionId) ?? []),
  1822			onSessionChange: (sessionId) => {
  1823				for (const key of observedBySession.keys()) if (key !== sessionId) observedBySession.delete(key);
  1824			}
  1825		});
  1826		if (!handle) return void 0;
  1827		current = {
  1828			handle,
  1829			references: 0
  1830		};
  1831		return leaseCurrent();
  1832	};
  1833	//#endregion
  1834	//#region src/client/replay-loader.ts
  1835	const noOpHandle = { stop: () => Promise.resolve() };
  1836	/**
  1837	* Own the complete client-session lifecycle. Consent can change after startup:
  1838	* grant starts capture, revoke synchronously detaches producers and discards
  1839	* buffers, and a later grant creates a clean runtime.
  1840	*/
  1841	const startClientSession = (config) => {
  1842		configurePrivacy(config.privacy);
  1843		if (!hasConsent()) clearPendingEvents();
  1844		const ingestKey = config.ingestKey;
  1845		if (typeof window === "undefined" || typeof document === "undefined" || !ingestKey || readSessionSink()) return noOpHandle;
  1846		const engineConfig = {
  1847			endpoint: config.endpoint.replace(/\/$/, ""),
  1848			ingestKey,
  1849			sdk: CLIENT_SDK_HINT,
  1850			maskAllInputs: config.replay?.maskAllInputs ?? true,
  1851			maskAllText: config.replay?.maskAllText ?? false,
  1852			getIdentity: getCurrentIdentity
  1853		};
  1854		const sampled = (config.replay?.enabled ?? true) && typeof document !== "undefined" && Math.random() < (config.replay?.sampleRate ?? 1);
  1855		let runtime;
  1856		let stopped = false;
  1857		let generation = 0;
  1858		let rotateOnNextStart = false;
  1859		const stopRuntime = async (flush) => {
  1860			generation++;
  1861			const previous = runtime;
  1862			runtime = void 0;
  1863			if (!previous) return;
  1864			const replayShutdown = previous.replay?.shutdown({ flush });
  1865			const metadataShutdown = previous.metadata?.shutdown({ flush });
  1866			const currentSessionId = previous.replay?.sessionId ?? previous.metadata?.sessionId;
  1867			const liveSink = getActiveSink();
  1868			const sink = liveSink && currentSessionId && liveSink.sessionId === currentSessionId ? liveSink : previous.sink;
  1869			if (flush) await sink.flush(true);
  1870			sink.stop();
  1871			if (sink !== previous.sink) previous.sink.stop();
  1872			if (currentSessionId) clearSessionSink(currentSessionId);
  1873			await Promise.all([replayShutdown, metadataShutdown]);
  1874		};
  1875		const startRuntime = () => {
  1876			if (stopped || runtime || !hasConsent() || readSessionSink()) return;
  1877			setVisitorTracking((config.privacy?.persistVisitorId ?? true) && mayPersistIdentifier());
  1878			const session = (rotateOnNextStart ? rotateSession() : void 0) ?? getSession();
  1879			rotateOnNextStart = false;
  1880			const next = { sink: startEventSink(engineConfig, session.id) };
  1881			runtime = next;
  1882			const ownGeneration = ++generation;
  1883			if (sampled) {
  1884				import("./replay-DBMTP4cB.mjs").then(({ startReplaySession }) => {
  1885					if (stopped || !hasConsent() || generation !== ownGeneration || runtime !== next) return;
  1886					next.replay = startReplaySession({
  1887						endpoint: config.endpoint,
  1888						ingestKey,
  1889						sdk: CLIENT_SDK_HINT,
  1890						serviceName: config.serviceName,
  1891						environment: config.environment,
  1892						serviceVersion: config.serviceVersion,
  1893						maskAllInputs: engineConfig.maskAllInputs,
  1894						maskAllText: engineConfig.maskAllText,
  1895						getIdentity: getCurrentIdentity,
  1896						captureUserEmail: config.privacy?.captureUserEmail
  1897					});
  1898				}).catch(() => {
  1899					if ((config.emitSessionMeta ?? true) && !stopped && hasConsent() && generation === ownGeneration && runtime === next) next.metadata = setupStandaloneSession({
  1900						endpoint: config.endpoint,
  1901						ingestKey: config.ingestKey,
  1902						serviceName: config.serviceName,
  1903						environment: config.environment,
  1904						serviceVersion: config.serviceVersion,
  1905						captureUserEmail: config.privacy?.captureUserEmail
  1906					});
  1907				});
  1908				return;
  1909			}
  1910			if (config.emitSessionMeta ?? true) next.metadata = setupStandaloneSession({
  1911				endpoint: config.endpoint,
  1912				ingestKey: config.ingestKey,
  1913				serviceName: config.serviceName,
  1914				environment: config.environment,
  1915				serviceVersion: config.serviceVersion,
  1916				captureUserEmail: config.privacy?.captureUserEmail
  1917			});
  1918		};
  1919		startRuntime();
  1920		const stopConsentListener = config.privacy?.requireConsent ? onConsentChange((allowed) => {
  1921			if (allowed) {
  1922				startRuntime();
  1923				return;
  1924			}
  1925			rotateOnNextStart = runtime !== void 0;
  1926			clearPendingEvents();
  1927			setVisitorTracking(false);
  1928			stopRuntime(false);
  1929		}) : () => {};
  1930		return { stop: async (options) => {
  1931			if (stopped) return;
  1932			stopped = true;
  1933			stopConsentListener();
  1934			await stopRuntime(options?.flush ?? true);
  1935		} };
  1936	};
  1937	//#endregion

@@@@@@@@@@ END FILE: sdk-excerpts/07-client-session-startup.txt

@@@@@@@@@@ FILE: sdk-excerpts/08-replay-record-final-chunk.txt (13459 bytes)
Replay engine: console/fetch/XHR wrapping, recorder flush (async gzip before POST), onSuspend
=============================================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/replay-DBMTP4cB.mjs  lines 1-145  (console + network capture + event capture) ---
     1	import { _ as BLOCK_SELECTOR, c as publishSessionSink, d as markActivity, f as nextChunkSeq, g as warnDropped, h as postSessionMeta, l as startSessionLifecycle, m as postSessionBlob, p as gzip, s as getObservedTraceIds, u as startEventSink, v as safeEmit, y as activeTraceId } from "./client-BYIixzTb.mjs";
     2	//#region ../browser-session/src/replay/capture/console.ts
     3	const LEVELS = [
     4		"log",
     5		"info",
     6		"warn",
     7		"error",
     8		"debug"
     9	];
    10	const MAX_MESSAGE = 2e3;
    11	/**
    12	* Capture `console.*` calls as session events. Wraps each method, emits a
    13	* distilled record, then forwards to the original so the host app's console
    14	* behaves normally. Never throws into the call site.
    15	*/
    16	function installConsoleCapture(emit) {
    17		const original = {};
    18		for (const level of LEVELS) {
    19			const orig = console[level];
    20			original[level] = orig;
    21			console[level] = (...args) => {
    22				safeEmit(emit, {
    23					type: "console",
    24					level,
    25					message: formatArgs(args)
    26				});
    27				orig.apply(console, args);
    28			};
    29		}
    30		return () => {
    31			for (const level of LEVELS) {
    32				const orig = original[level];
    33				if (orig) console[level] = orig;
    34			}
    35		};
    36	}
    37	function formatArgs(args) {
    38		const text = args.map((a) => {
    39			if (typeof a === "string") return a;
    40			if (a instanceof Error) return `${a.name}: ${a.message}`;
    41			try {
    42				return JSON.stringify(a);
    43			} catch {
    44				return String(a);
    45			}
    46		}).join(" ");
    47		return text.length > MAX_MESSAGE ? `${text.slice(0, MAX_MESSAGE)}…` : text;
    48	}
    49	//#endregion
    50	//#region ../browser-session/src/replay/capture/network.ts
    51	/**
    52	* Capture fetch + XHR requests as session events, tagged with the active trace
    53	* id so each request links to its backend trace. `ignoreUrl` skips Maple's own
    54	* ingest endpoints (otherwise capturing the session-events POST would loop).
    55	*/
    56	function installNetworkCapture(emit, ignoreUrl) {
    57		const origFetch = typeof window !== "undefined" ? window.fetch : void 0;
    58		if (origFetch) window.fetch = async (input, init) => {
    59			const url = requestUrl(input);
    60			const method = requestMethod(input, init);
    61			const traceId = activeTraceId();
    62			const start = performance.now();
    63			try {
    64				const res = await origFetch(input, init);
    65				record(url, method, res.status, start, traceId);
    66				return res;
    67			} catch (error) {
    68				record(url, method, 0, start, traceId, String(error));
    69				throw error;
    70			}
    71		};
    72		const record = (url, method, status, start, traceId, error) => {
    73			if (ignoreUrl(url)) return;
    74			safeEmit(emit, {
    75				type: "network",
    76				net: {
    77					method,
    78					url,
    79					status,
    80					durationMs: Math.round(performance.now() - start)
    81				},
    82				traceId,
    83				...error ? { attrs: { error } } : void 0
    84			});
    85		};
    86		const XHR = typeof window !== "undefined" ? window.XMLHttpRequest : void 0;
    87		const origOpen = XHR?.prototype.open;
    88		const origSend = XHR?.prototype.send;
    89		if (XHR && origOpen && origSend) {
    90			XHR.prototype.open = function(method, url, ...rest) {
    91				this.__mapleMethod = String(method).toUpperCase();
    92				this.__mapleUrl = typeof url === "string" ? url : url.href;
    93				return origOpen.apply(this, [
    94					method,
    95					url,
    96					...rest
    97				]);
    98			};
    99			XHR.prototype.send = function(...args) {
   100				const meta = this;
   101				const start = performance.now();
   102				const traceId = activeTraceId();
   103				this.addEventListener("loadend", () => {
   104					record(meta.__mapleUrl ?? "", meta.__mapleMethod ?? "GET", this.status, start, traceId);
   105				});
   106				return origSend.apply(this, args);
   107			};
   108		}
   109		return () => {
   110			if (origFetch) window.fetch = origFetch;
   111			if (XHR && origOpen) XHR.prototype.open = origOpen;
   112			if (XHR && origSend) XHR.prototype.send = origSend;
   113		};
   114	}
   115	function requestUrl(input) {
   116		if (typeof input === "string") return input;
   117		if (input instanceof URL) return input.href;
   118		return input.url;
   119	}
   120	function requestMethod(input, init) {
   121		return (init?.method ?? (typeof input === "object" && "method" in input ? input.method : void 0) ?? "GET").toUpperCase();
   122	}
   123	//#endregion
   124	//#region ../browser-session/src/replay/events.ts
   125	/**
   126	* Install the capture modules that turn console output and network requests
   127	* into distilled session events.
   128	*
   129	* Navigation, errors and interactions are deliberately absent: the sink
   130	* installs those itself (see `startBaselineCapture`), because page views, error
   131	* counts and click counts have to be counted even when replay is unsampled —
   132	* and a second listener here would double-count every one of them.
   133	*/
   134	function startEventCapture(config, sessionId) {
   135		const sink = startEventSink(config, sessionId);
   136		const emit = sink.emit;
   137		const uninstall = [installConsoleCapture(emit), installNetworkCapture(emit, sink.ignoreUrl)];
   138		return {
   139			stop: () => {
   140				for (const off of uninstall) off();
   141			},
   142			flush: sink.flush
   143		};
   144	}
   145	//#endregion

--- @maple-dev/effect-sdk/dist/replay-DBMTP4cB.mjs  lines 10591-10779  (../browser-session/src/replay/record.ts + session/replay-session.ts) ---
 10591	//#region ../browser-session/src/replay/record.ts
 10592	const FULL_SNAPSHOT = 2;
 10593	const INCREMENTAL = 3;
 10594	const SOURCE_MOUSE_INTERACTION = 2;
 10595	const MOUSE_CLICK = 2;
 10596	const FLUSH_INTERVAL_MS = 5e3;
 10597	const FLUSH_BYTES = 102400;
 10598	const CHECKOUT_EVERY_MS = 3e5;
 10599	const MAX_BUFFER_BYTES = 4194304;
 10600	function warnExhausted(sessionId) {
 10601		console.warn(`[maple] session replay ${sessionId} reached its maximum recorded size; recording stopped for this session (metadata and events continue)`);
 10602	}
 10603	let lastDropWarnAt = 0;
 10604	function warnBufferDropped(bytes) {
 10605		const now = Date.now();
 10606		if (now - lastDropWarnAt < 3e4) return;
 10607		lastDropWarnAt = now;
 10608		console.warn(`[maple] session replay buffer exceeded ${MAX_BUFFER_BYTES} bytes (dropping ${bytes} buffered bytes; recording continues from the next full snapshot)`);
 10609	}
 10610	function startRecording(config, sessionId) {
 10611		let parts = [];
 10612		let bufferBytes = 0;
 10613		let bufferHasCheckpoint = false;
 10614		let firstTimestamp = 0;
 10615		let lastTimestamp = 0;
 10616		let droppedChunk = false;
 10617		let clickCount = 0;
 10618		let stopped = false;
 10619		let exhausted = false;
 10620		const resetBuffer = () => {
 10621			parts = [];
 10622			bufferBytes = 0;
 10623			bufferHasCheckpoint = false;
 10624			firstTimestamp = 0;
 10625			lastTimestamp = 0;
 10626		};
 10627		const flush = async (keepalive = false) => {
 10628			if (stopped || exhausted || parts.length === 0) return;
 10629			const body = `[${parts.join(",")}]`;
 10630			const isCheckpoint = bufferHasCheckpoint;
 10631			const eventCount = parts.length;
 10632			const durationMs = Math.max(0, lastTimestamp - firstTimestamp);
 10633			const seq = nextChunkSeq();
 10634			resetBuffer();
 10635			let gzipped;
 10636			try {
 10637				gzipped = await gzip(new TextEncoder().encode(body));
 10638			} catch (error) {
 10639				warnDropped("chunk compression", error);
 10640				return;
 10641			}
 10642			if (await postSessionBlob(config, {
 10643				sessionId,
 10644				chunkSeq: seq,
 10645				isCheckpoint,
 10646				eventCount,
 10647				durationMs
 10648			}, gzipped, keepalive) === "exhausted" && !exhausted) {
 10649				exhausted = true;
 10650				warnExhausted(sessionId);
 10651				haltCapture();
 10652			}
 10653		};
 10654		let haltCapture = () => {};
 10655		const stop = record({
 10656			emit: (event, isCheckpoint) => {
 10657				const active = markActivity();
 10658				if (active && active.id !== sessionId) return;
 10659				const e = event;
 10660				const isFullSnapshot = isCheckpoint === true || e.type === FULL_SNAPSHOT;
 10661				if (e.type === INCREMENTAL && e.data?.source === SOURCE_MOUSE_INTERACTION && e.data.type === MOUSE_CLICK) clickCount++;
 10662				let json;
 10663				try {
 10664					json = JSON.stringify(e);
 10665				} catch {
 10666					return;
 10667				}
 10668				if (droppedChunk && !isFullSnapshot) return;
 10669				droppedChunk = false;
 10670				if (bufferBytes + json.length > MAX_BUFFER_BYTES) {
 10671					warnBufferDropped(bufferBytes + json.length);
 10672					resetBuffer();
 10673					droppedChunk = true;
 10674					return;
 10675				}
 10676				if (isFullSnapshot) bufferHasCheckpoint = true;
 10677				if (parts.length === 0) firstTimestamp = e.timestamp;
 10678				lastTimestamp = e.timestamp;
 10679				parts.push(json);
 10680				bufferBytes += json.length;
 10681				if (bufferBytes >= FLUSH_BYTES) flush();
 10682			},
 10683			maskAllInputs: config.maskAllInputs,
 10684			blockSelector: BLOCK_SELECTOR,
 10685			...config.maskAllText ? { maskTextSelector: "*" } : void 0,
 10686			checkoutEveryNms: CHECKOUT_EVERY_MS
 10687		});
 10688		let idleHandle;
 10689		const scheduleFlush = () => {
 10690			if (typeof requestIdleCallback === "function") idleHandle = requestIdleCallback(() => {
 10691				idleHandle = void 0;
 10692				flush();
 10693			}, { timeout: 2e3 });
 10694			else flush();
 10695		};
 10696		const flushTimer = setInterval(scheduleFlush, FLUSH_INTERVAL_MS);
 10697		let halted = false;
 10698		haltCapture = () => {
 10699			if (halted) return;
 10700			halted = true;
 10701			clearInterval(flushTimer);
 10702			if (idleHandle !== void 0 && typeof cancelIdleCallback === "function") {
 10703				cancelIdleCallback(idleHandle);
 10704				idleHandle = void 0;
 10705			}
 10706			resetBuffer();
 10707			stop?.();
 10708		};
 10709		return {
 10710			stop: () => {
 10711				stopped = true;
 10712				haltCapture();
 10713			},
 10714			flush,
 10715			getClickCount: () => clickCount
 10716		};
 10717	}
 10718	//#endregion
 10719	//#region ../browser-session/src/session/replay-session.ts
 10720	/**
 10721	* Start recording the current browser session. Publishes the session sink,
 10722	* posts an `active` metadata row, and installs visibility handlers:
 10723	* hidden → flush + `ended` row (with observed trace ids) + stop capture;
 10724	* visible → re-resolve the session (rotating if idle-expired), republish the
 10725	* sink, restart capture, post a fresh `active` row. Metadata versions are
 10726	* monotonic per session, so the latest row always wins on the backend.
 10727	*
 10728	* Returns undefined outside a browser. Sampling is the caller's decision.
 10729	*/
 10730	function startReplaySession(options) {
 10731		if (typeof window === "undefined") return void 0;
 10732		const engineConfig = {
 10733			endpoint: options.endpoint.replace(/\/$/, ""),
 10734			ingestKey: options.ingestKey,
 10735			sdk: options.sdk,
 10736			maskAllInputs: options.maskAllInputs,
 10737			maskAllText: options.maskAllText,
 10738			getIdentity: options.getIdentity
 10739		};
 10740		let recorder;
 10741		let events;
 10742		let publishedSessionId;
 10743		const publish = (sessionId) => {
 10744			if (publishedSessionId === sessionId) return;
 10745			publishedSessionId = sessionId;
 10746			publishSessionSink(sessionId);
 10747		};
 10748		return startSessionLifecycle({
 10749			...options,
 10750			getTraceIds: getObservedTraceIds
 10751		}, {
 10752			recorded: true,
 10753			post: (row, keepalive) => {
 10754				postSessionMeta(engineConfig, row, keepalive);
 10755			},
 10756			clicksSinceStart: () => recorder?.getClickCount() ?? 0,
 10757			onStart: (record) => {
 10758				publish(record.id);
 10759				recorder = startRecording(engineConfig, record.id);
 10760				events = startEventCapture(engineConfig, record.id);
 10761			},
 10762			onSuspend: ({ flush, keepalive }) => {
 10763				const stoppingRecorder = recorder;
 10764				const stoppingEvents = events;
 10765				recorder = void 0;
 10766				events = void 0;
 10767				const flushed = flush ? Promise.all([stoppingRecorder?.flush(keepalive), stoppingEvents?.flush(keepalive)]).then(() => {}) : void 0;
 10768				stoppingRecorder?.stop();
 10769				stoppingEvents?.stop();
 10770				return flushed;
 10771			},
 10772			onSessionChange: (sessionId) => {
 10773				publish(sessionId);
 10774				options.onSessionChange?.(sessionId);
 10775			}
 10776		});
 10777	}
 10778	//#endregion
 10779	export { startReplaySession };

@@@@@@@@@@ END FILE: sdk-excerpts/08-replay-record-final-chunk.txt

@@@@@@@@@@ FILE: sdk-excerpts/09-maple-layer-client.txt (6922 bytes)
Maple.layer (client): Otlp.layerJson + consent HTTP client; no unload flush
===========================================================================

Verbatim excerpt from the published npm package @maple-dev/effect-sdk@0.8.2 (installed with `bun install` in ../repro).
Line numbers are those of the installed dist file. Nothing has been edited.

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 2177-2228  (src/client/consent-http-client.ts) ---
  2177	//#region src/client/consent-http-client.ts
  2178	const after = (value, threshold) => trySyncOrUndefined(() => BigInt(String(value ?? 0)) >= threshold) ?? false;
  2179	const pruneTraces = (body, threshold) => {
  2180		const resourceSpans = (body.resourceSpans ?? []).map((resource) => ({
  2181			...resource,
  2182			scopeSpans: (resource.scopeSpans ?? []).map((scope) => ({
  2183				...scope,
  2184				spans: (scope.spans ?? []).filter((span) => after(span.startTimeUnixNano, threshold))
  2185			})).filter((scope) => scope.spans.length > 0)
  2186		})).filter((resource) => resource.scopeSpans.length > 0);
  2187		return resourceSpans.length > 0 ? {
  2188			...body,
  2189			resourceSpans
  2190		} : void 0;
  2191	};
  2192	const pruneLogs = (body, threshold) => {
  2193		const resourceLogs = (body.resourceLogs ?? []).map((resource) => ({
  2194			...resource,
  2195			scopeLogs: (resource.scopeLogs ?? []).map((scope) => ({
  2196				...scope,
  2197				logRecords: (scope.logRecords ?? []).filter((record) => after(record.timeUnixNano ?? record.observedTimeUnixNano, threshold))
  2198			})).filter((scope) => scope.logRecords.length > 0)
  2199		})).filter((resource) => resource.scopeLogs.length > 0);
  2200		return resourceLogs.length > 0 ? {
  2201			...body,
  2202			resourceLogs
  2203		} : void 0;
  2204	};
  2205	/**
  2206	* Keeps signals captured under the current grant. Cumulative Effect metrics
  2207	* cannot be separated here, so explicit-consent installs suppress them.
  2208	*/
  2209	const filterOtlpRequestForConsent = (request, requireConsent) => {
  2210		if (!hasConsent()) return void 0;
  2211		if (!requireConsent) return request;
  2212		if (request.url.endsWith("/v1/metrics")) return void 0;
  2213		if (request.body._tag !== "Uint8Array") return void 0;
  2214		const body = request.body;
  2215		const since = consentAllowedSince();
  2216		if (!Number.isFinite(since)) return void 0;
  2217		return trySyncOrUndefined(() => {
  2218			const decoded = JSON.parse(new TextDecoder().decode(body.body));
  2219			const threshold = BigInt(Math.trunc(since)) * 1000000n;
  2220			const filtered = request.url.endsWith("/v1/traces") ? pruneTraces(decoded, threshold) : request.url.endsWith("/v1/logs") ? pruneLogs(decoded, threshold) : void 0;
  2221			return filtered ? HttpClientRequest.setBody(request, HttpBody.jsonUnsafe(filtered, body.contentType)) : void 0;
  2222		});
  2223	};
  2224	const consentHttpClientLayer = (requireConsent) => Layer.effect(HttpClient.HttpClient, Effect.map(HttpClient.HttpClient, (inner) => HttpClient.make((request) => {
  2225		const filtered = filterOtlpRequestForConsent(request, requireConsent);
  2226		return filtered ? inner.execute(filtered) : Effect.succeed(HttpClientResponse.fromWeb(request, new Response(null, { status: 200 })));
  2227	}))).pipe(Layer.provide(FetchHttpClient.layer));
  2228	//#endregion

--- @maple-dev/effect-sdk/dist/client-BYIixzTb.mjs  lines 2257-2330  (src/client/layer.ts) ---
  2257	//#region src/client/layer.ts
  2258	var layer_exports = /* @__PURE__ */ __exportAll({
  2259		clearIdentity: () => clearIdentity,
  2260		identify: () => identify,
  2261		layer: () => layer,
  2262		track: () => track
  2263	});
  2264	/**
  2265	* Create an Effect Layer that provides OpenTelemetry traces, logs, and metrics
  2266	* configured for Maple in browser environments.
  2267	*
  2268	* Unlike the server layer, all configuration must be provided programmatically
  2269	* since browsers don't have access to environment variables.
  2270	*
  2271	* @example
  2272	* ```typescript
  2273	* import { Maple } from "@maple-dev/effect-sdk/client"
  2274	* import { Effect } from "effect"
  2275	*
  2276	* const TracerLive = Maple.layer({
  2277	*   serviceName: "my-frontend",
  2278	*   endpoint: "https://ingest.maple.dev",
  2279	*   ingestKey: "maple_pk_...",
  2280	* })
  2281	*
  2282	* const program = Effect.log("Hello!").pipe(Effect.withSpan("hello"))
  2283	* Effect.runPromise(program.pipe(Effect.provide(TracerLive)))
  2284	* ```
  2285	*/
  2286	const layer = (config) => {
  2287		const attributes = { "maple.sdk.type": "client" };
  2288		const nav = browserNavigator();
  2289		if (nav) {
  2290			if (nav.userAgent) attributes["user_agent.original"] = nav.userAgent;
  2291			if (nav.language) attributes["browser.language"] = nav.language;
  2292		}
  2293		if (typeof Intl !== "undefined") {
  2294			const timezone = trySyncOrUndefined(() => Intl.DateTimeFormat().resolvedOptions().timeZone);
  2295			if (timezone) attributes["browser.timezone"] = timezone;
  2296		}
  2297		if (config.environment) {
  2298			attributes["deployment.environment"] = config.environment;
  2299			attributes["deployment.environment.name"] = config.environment;
  2300		}
  2301		if (config.serviceVersion && /^[0-9a-f]{7,40}$/i.test(config.serviceVersion)) attributes["vcs.ref.head.revision"] = config.serviceVersion;
  2302		if (config.serviceNamespace) attributes["service.namespace"] = config.serviceNamespace;
  2303		if (config.attributes) Object.assign(attributes, config.attributes);
  2304		const clientSessionConfig = {
  2305			endpoint: config.endpoint,
  2306			ingestKey: config.ingestKey,
  2307			serviceName: config.serviceName,
  2308			environment: config.environment,
  2309			serviceVersion: config.serviceVersion,
  2310			replay: config.replay,
  2311			emitSessionMeta: config.emitSessionMeta,
  2312			privacy: config.privacy
  2313		};
  2314		const base = Otlp.layerJson({
  2315			baseUrl: config.endpoint,
  2316			resource: {
  2317				serviceName: config.serviceName,
  2318				serviceVersion: config.serviceVersion,
  2319				attributes
  2320			},
  2321			headers: config.ingestKey ? { Authorization: `Bearer ${config.ingestKey}` } : void 0,
  2322			maxBatchSize: config.maxBatchSize,
  2323			loggerExportInterval: config.loggerExportInterval,
  2324			metricsExportInterval: config.metricsExportInterval,
  2325			tracerExportInterval: config.tracerExportInterval,
  2326			shutdownTimeout: config.shutdownTimeout
  2327		}).pipe(Layer.provide(consentHttpClientLayer(config.privacy?.requireConsent ?? false)));
  2328		const lifecycle = Layer.effectDiscard(Effect.acquireRelease(Effect.sync(() => startClientSession(clientSessionConfig)), (clientSession) => Effect.promise(() => clientSession.stop().catch(() => void 0))));
  2329		return withSessionLink(Layer.merge(base, lifecycle));
  2330	};

@@@@@@@@@@ END FILE: sdk-excerpts/09-maple-layer-client.txt

@@@@@@@@@@ FILE: sdk-excerpts/10-effect-otlp-exporter.txt (8426 bytes)
Effect (NOT Maple) OtlpExporter used by Maple.layer: Flusher service, 60 s disable + drop on failure
====================================================================================================

Verbatim excerpt from effect@4.0.0-rc.112 (the SDK's Effect peer), installed in ../repro. Included because Maple.layer delegates export to it.

--- effect/dist/unstable/observability/OtlpExporter.js  lines 41-181  (effect@4.0.0-rc.112) ---
    41	const policy = /*#__PURE__*/Schedule.forever.pipe(Schedule.passthrough, /*#__PURE__*/Schedule.addDelay(({
    42	  output: error
    43	}) => {
    44	  if (HttpClientError.isHttpClientError(error) && error.reason._tag === "StatusCodeError" && error.reason.response.status === 429) {
    45	    return retryAfterDelay(error.reason.response.headers["retry-after"]);
    46	  }
    47	  return Effect.succeed(Duration.seconds(1));
    48	}));
    49	/**
    50	 * Registry of exporter flush operations, used to manually drain buffered
    51	 * telemetry before the surrounding scope closes.
    52	 *
    53	 * **Details**
    54	 *
    55	 * Every exporter created by `make` registers its export operation here, so a
    56	 * single `flush` drains all signals sharing the registry. `flush` returns only
    57	 * after the exports it initiated have settled, cannot fail, and respects each
    58	 * exporter's temporary-disable window. Wrap it with `Effect.timeoutOption` to
    59	 * bound its duration at the call site.
    60	 *
    61	 * @category services
    62	 * @since 4.0.0
    63	 */
    64	export class Flusher extends /*#__PURE__*/Context.Service()("effect/observability/OtlpExporter/Flusher") {}
    65	/**
    66	 * Provides a `Flusher` backed by a fresh registry.
    67	 *
    68	 * **Details**
    69	 *
    70	 * This is intentionally a single module-level constant rather than a factory:
    71	 * layer memoization is keyed by layer instance, so every signal layer
    72	 * referencing this same constant shares one registry per layer build, and one
    73	 * `flush` drains traces, logs and metrics together. A factory returning a new
    74	 * layer per call would silently create one registry per signal.
    75	 *
    76	 * Registration is scoped — an exporter is removed from the registry when its
    77	 * own scope closes. Flushing with an empty registry is a no-op.
    78	 *
    79	 * Note that `flush` cannot await an export that was already in flight when it
    80	 * was called (for example one started by the export interval); it only waits
    81	 * for the exports it initiates.
    82	 *
    83	 * @category layers
    84	 * @since 4.0.0
    85	 */
    86	export const layerFlusher = /*#__PURE__*/Layer.sync(Flusher, () => {
    87	  const registry = new Set();
    88	  return {
    89	    flush: Effect.suspend(() => {
    90	      if (registry.size === 0) {
    91	        return Effect.void;
    92	      }
    93	      return Effect.forEach(registry, identity, {
    94	        concurrency: "unbounded",
    95	        discard: true
    96	      });
    97	    }),
    98	    register: run => Effect.flatMap(Scope.Scope, scope => {
    99	      registry.add(run);
   100	      return Scope.addFinalizer(scope, Effect.sync(() => registry.delete(run)));
   101	    })
   102	  };
   103	});
   104	/**
   105	 * Creates a scoped OTLP batch exporter.
   106	 *
   107	 * **Details**
   108	 *
   109	 * The exporter buffers pushed data, periodically posts encoded batches to the
   110	 * configured URL, retries transient failures, temporarily disables exporting
   111	 * after unhandled failures, and flushes during scope finalization up to
   112	 * `shutdownTimeout`.
   113	 *
   114	 * @category constructors
   115	 * @since 4.0.0
   116	 */
   117	export const make = /*#__PURE__*/Effect.fnUntraced(function* (options) {
   118	  const services = yield* Effect.context();
   119	  const clock = Context.get(services, Clock);
   120	  const scope = Context.get(services, Scope.Scope);
   121	  const exportInterval = Duration.max(Duration.fromInputUnsafe(options.exportInterval), Duration.zero);
   122	  let disabledUntil = undefined;
   123	  const client = HttpClient.filterStatusOk(Context.get(services, HttpClient.HttpClient)).pipe(HttpClient.transformResponse(Effect.provideService(HttpClient.TracerPropagationEnabled, false)), HttpClient.retryTransient({
   124	    schedule: policy,
   125	    times: 3
   126	  }));
   127	  let headers = Headers.fromRecordUnsafe({
   128	    "user-agent": `effect-opentelemetry-${options.label}/0.0.0`
   129	  });
   130	  if (options.headers) {
   131	    headers = Headers.merge(Headers.fromInput(options.headers), headers);
   132	  }
   133	  const request = HttpClientRequest.post(options.url, {
   134	    headers
   135	  });
   136	  let buffer = [];
   137	  const runExport = Effect.suspend(() => {
   138	    if (disabledUntil !== undefined && clock.currentTimeMillisUnsafe() < disabledUntil) {
   139	      return Effect.void;
   140	    } else if (disabledUntil !== undefined) {
   141	      disabledUntil = undefined;
   142	    }
   143	    const items = buffer;
   144	    if (options.maxBatchSize !== "disabled") {
   145	      if (buffer.length === 0) {
   146	        return Effect.void;
   147	      }
   148	      buffer = [];
   149	    }
   150	    const [body, onSuccess] = options.body(items);
   151	    return client.execute(HttpClientRequest.setBody(request, body)).pipe(Effect.andThen(onSuccess), Effect.asVoid, Effect.withTracerEnabled(false));
   152	  }).pipe(Effect.catchCause(cause => {
   153	    if (disabledUntil !== undefined) return Effect.void;
   154	    disabledUntil = clock.currentTimeMillisUnsafe() + 60_000;
   155	    buffer = [];
   156	    return Effect.logDebug("Disabling exporter for 60 seconds", cause);
   157	  }), Effect.annotateLogs({
   158	    package: "@effect/opentelemetry",
   159	    module: options.label
   160	  }));
   161	  const exportFibers = yield* FiberSet.make();
   162	  const runExportFork = yield* FiberSet.runtime(exportFibers)();
   163	  const flusher = yield* Flusher;
   164	  yield* flusher.register(runExport);
   165	  yield* Scope.addFinalizer(scope, Effect.suspend(() => {
   166	    if (disabledUntil !== undefined) return Effect.void;
   167	    runExportFork(runExport);
   168	    return FiberSet.awaitEmpty(exportFibers);
   169	  }).pipe(Effect.ignore, Effect.interruptible, Effect.timeoutOption(options.shutdownTimeout)));
   170	  yield* Effect.sleep(exportInterval).pipe(Effect.andThen(FiberSet.run(exportFibers, runExport)), Effect.flatMap(Fiber.await), Effect.forever, Effect.forkIn(scope));
   171	  return {
   172	    push(data) {
   173	      if (disabledUntil !== undefined) return;
   174	      buffer.push(data);
   175	      if (options.maxBatchSize !== "disabled" && buffer.length >= options.maxBatchSize) {
   176	        runExportFork(runExport);
   177	      }
   178	    }
   179	  };
   180	});
   181	//# sourceMappingURL=OtlpExporter.js.map

--- effect/dist/unstable/observability/OtlpTracer.js  lines 94-94  (OtlpTracer.layer provides layerFlusher) ---
    94	export const layer = /*#__PURE__*/flow(make, /*#__PURE__*/Layer.effect(Tracer.Tracer), /*#__PURE__*/Layer.provideMerge(Exporter.layerFlusher));

--- effect/dist/unstable/http/FetchHttpClient.js  lines 40-60  (FetchHttpClient.RequestInit service) ---
    40	 *
    41	 * Request-specific method, headers, body, and abort signal are supplied by the client when a request is executed.
    42	 *
    43	 * @category services
    44	 * @since 4.0.0
    45	 */
    46	export class RequestInit extends /*#__PURE__*/Context.Service()("effect/http/FetchHttpClient/RequestInit") {}
    47	const fetch = /*#__PURE__*/HttpClient.make((request, url, signal, fiber) => {
    48	  const fetch = fiber.getRef(Fetch);
    49	  const options = Context.getOrUndefined(fiber.context, RequestInit) ?? {};
    50	  let headers = options.headers ? Headers.merge(Headers.fromInput(options.headers), request.headers) : request.headers;
    51	  if (headers["content-length"]) {
    52	    headers = Headers.remove(headers, "content-length");
    53	  }
    54	  const send = body => Effect.map(Effect.tryPromise({
    55	    try: () => fetch(url, {
    56	      ...options,
    57	      method: request.method,
    58	      headers,
    59	      body,
    60	      duplex: request.body._tag === "Stream" ? "half" : undefined,

@@@@@@@@@@ END FILE: sdk-excerpts/10-effect-otlp-exporter.txt

@@@@@@@@@@ FILE: sdk-excerpts/11-readme.txt (5172 bytes)
README.md of @maple-dev/effect-sdk@0.8.2: session replay + manual flush sections
================================================================================

Verbatim excerpt from the package README.

--- @maple-dev/effect-sdk/README.md  lines 156-177  (Session replay & sessions) ---
   156	### Session replay & sessions (built in)
   157	
   158	The browser presets (`Maple.layer` and `MapleFlush.make`) record **rrweb session replays by default** — no separate browser SDK needed. Every span carries a `session.id`, the session appears in Maple's Sessions UI with its linked traces, and the recording is playable next to them.
   159	
   160	```typescript
   161	const TracerLive = Maple.layer({
   162		serviceName: "my-frontend",
   163		endpoint: "https://ingest.maple.dev",
   164		ingestKey: "maple_pk_...",
   165		replay: {
   166			sampleRate: 0.1, // record 10% of sessions (default 1)
   167			maskAllInputs: true, // default
   168			maskAllText: false, // default
   169		},
   170	})
   171	```
   172	
   173	- **Bundle size:** the replay engine (rrweb included) loads through a dynamic import, so it lands in a code-split chunk (~360 kB) fetched only when replay is enabled _and_ the session is sampled. The base client bundle stays ~13 kB.
   174	- **Vendored dependency:** rrweb (MIT) is bundled into that chunk rather than declared in `dependencies`, so it will not appear in your lockfile or in `npm ls` output. Note it directly if your license or supply-chain audit enumerates transitive packages.
   175	- **Opt out** with `replay: { enabled: false }`. Unsampled or disabled sessions still appear in the Sessions UI (metadata rows + linked traces, no recording); turn that off too with `emitSessionMeta: false`.
   176	- **Tab lifecycle:** recording suspends on `visibilitychange → hidden` (flushing the tail with `keepalive`) and resumes when the tab becomes visible again, rotating to a fresh session after 30 minutes of inactivity.
   177	- **Identify users** at any point — the id is attached to the session's next-posted metadata row _and_ stamped as `user.id` on every span the client tracer creates from then on (traces become user-attributable, not just session-grouped). Spans created before you call it stay anonymous. Pass `null` or `undefined` after sign-out to make future telemetry anonymous again.

--- @maple-dev/effect-sdk/README.md  lines 235-277  (Manual flush) ---
   235	## Manual flush
   236	
   237	`Maple.layer` (server + client) batches in the background and only exports on a timer, on batch overflow, or when its scope closes — there's no way to force an export. That's a problem in two places: a browser tab dropping the last few seconds of spans on unload, and a short-lived process exiting before the timer fires.
   238	
   239	`MapleFlush.make()` (available from both `/server` and `/client`) swaps the background exporter for the same buffer-backed tracer/logger the Cloudflare preset uses, and returns an explicit `flush()`:
   240	
   241	```typescript
   242	export interface FlushableTelemetry {
   243		readonly layer: Layer.Layer<never>
   244		readonly flush: () => Promise<void> // drain buffers → POST now (never rejects)
   245		readonly dispose: () => Promise<void> // stop the auto-flush timer/listeners + final flush
   246	}
   247	```
   248	
   249	Both presets run a background auto-flush every 5s by default (configurable via `autoFlushInterval`, or `false` to flush purely on demand), so it's a safe drop-in for `Maple.layer` with manual flush layered on top. Effect metrics are cumulative, but the flushable presets only post a new snapshot after a metric changes; an unchanged registry does not generate a metrics request on every flush tick.
   250	
   251	### Server / Node
   252	
   253	```typescript
   254	import { MapleFlush } from "@maple-dev/effect-sdk/server"
   255	
   256	const telemetry = MapleFlush.make({ serviceName: "my-app" }) // same env auto-detect as Maple.layer
   257	
   258	// ...provide telemetry.layer to your runtime...
   259	await telemetry.flush() // force an export at a checkpoint
   260	await telemetry.dispose() // before exit: stop the timer + final flush
   261	```
   262	
   263	### Client / Browser
   264	
   265	```typescript
   266	import { MapleFlush } from "@maple-dev/effect-sdk/client"
   267	
   268	const telemetry = MapleFlush.make({
   269		serviceName: "my-frontend",
   270		endpoint: "https://ingest.maple.dev",
   271		ingestKey: "maple_pk_...",
   272		// flushOnUnload: true (default) registers pagehide / visibilitychange→hidden handlers
   273	})
   274	// telemetry.layer keeps the replay-session trace linking from Maple.layer.
   275	```
   276	
   277	By default the client preset flushes on `pagehide` and `visibilitychange→hidden` so the tail of a session isn't lost when the tab goes away. Flush uses `fetch(url, { keepalive: true })`, **not** `navigator.sendBeacon`: Maple's ingest authenticates via the `Authorization` header (no query-param auth) and sendBeacon can't set headers, so it would 401 whenever an ingest key is set. `keepalive` carries the header and still survives unload for small bodies.

@@@@@@@@@@ END FILE: sdk-excerpts/11-readme.txt

@@@@@@@@@@ FILE: sdk-excerpts/12-v0.10.0-what-changed-and-what-did-not.txt (10614 bytes)
@maple-dev/effect-sdk@0.10.0 (latest at time of writing): shared keepalive budget for session posts; OTLP flush path unchanged
==============================================================================================================================

Verbatim excerpt from @maple-dev/effect-sdk@0.10.0 (npm, published 2026-10-03), installed in ../repro/variants/sdk-0.10.0 with effect@4.0.0. Shown for comparison only; the report's findings were made against 0.8.2 and re-checked against 0.10.0 where stated.

--- @maple-dev/effect-sdk/dist/version-B9ejEduZ.mjs  lines 156-245  (flushSignal / guardFlush / makeSerializedFlush / runFlush - cooldown, whole-batch POST, serialization unchanged in substance) ---
   156	const flushSignal = async (args) => {
   157		const { url, headers, buffer, body, state, signal, transport, logPrefix } = args;
   158		if (state.disabledUntil && Date.now() < state.disabledUntil) {
   159			console.warn(`${logPrefix} ${signal} flush skipped (cooldown ${state.disabledUntil - Date.now()}ms remaining)`);
   160			return;
   161		}
   162		state.disabledUntil = 0;
   163		const batch = buffer.drain();
   164		if (batch.length === 0) return;
   165		const posted = await Effect.runPromise(Effect.result(Effect.tryPromise({
   166			try: () => transport.post(url, headers, body(batch)),
   167			catch: (cause) => cause
   168		})));
   169		if (Result.isFailure(posted)) {
   170			buffer.restore(batch);
   171			state.disabledUntil = Date.now() + COOLDOWN_MS;
   172			console.error(`${logPrefix} ${signal} flush failed; cooldown 60s:`, posted.failure);
   173		}
   174	};
   175	/**
   176	* Wrap a flush body so it logs its failure rather than rejecting.
   177	*
   178	* Every flush is documented never to reject: they are fired as `void flush()`
   179	* from auto-flush timers, `pagehide`/`visibilitychange` handlers, and
   180	* `ctx.waitUntil`, where a rejection is an unhandled rejection — fatal under
   181	* `--unhandled-rejections=strict` — caused purely by telemetry. `runFlush`
   182	* already absorbs per-signal transport errors; this covers everything around it,
   183	* chiefly resource resolution.
   184	*/
   185	const guardFlush = (logPrefix, run) => async (...args) => {
   186		const outcome = await Effect.runPromise(Effect.result(Effect.tryPromise({
   187			try: () => run(...args),
   188			catch: (cause) => cause
   189		})));
   190		if (Result.isFailure(outcome)) console.error(`${logPrefix} flush failed:`, outcome.failure);
   191	};
   192	/**
   193	* Serialize drains. Workers may coalesce queued calls with identical arguments:
   194	* every caller waits for that drain, and calls arriving DURING it queue a later
   195	* drain so spans completed after the first snapshot cannot be stranded.
   196	* Other presets retain one drain per call by default.
   197	*/
   198	const makeSerializedFlush = (run, options) => {
   199		let tail = Promise.resolve();
   200		let queued;
   201		return (...args) => {
   202			const waiting = queued;
   203			if (options?.coalesceSameArguments && waiting && args.length === waiting.args.length && args.every((value, index) => value === waiting.args[index])) return waiting.promise;
   204			const next = tail.then(() => {
   205				if (queued?.promise === next) queued = void 0;
   206				return run(...args);
   207			});
   208			queued = {
   209				args,
   210				promise: next
   211			};
   212			tail = next.catch(() => void 0);
   213			return next;
   214		};
   215	};
   216	/**
   217	* Drain the span + log buffers and POST them. Errors are swallowed per signal
   218	* (see {@link flushSignal}); the returned promise never rejects.
   219	*
   220	* - `noOp`: drain so the buffers don't grow unbounded, fire `onNoOp` (one-shot
   221	*   "telemetry disabled" notice), never POST.
   222	* - empty buffers: short-circuit without a request.
   223	*/
   224	const runFlush = async (args) => {
   225		const { resolved: r, spans, logs, metrics, tracesState, logsState, metricsState, transport, logPrefix, onNoOp } = args;
   226		if (r.noOp) {
   227			spans.drain();
   228			logs.drain();
   229			metrics.drain();
   230			onNoOp?.();
   231			return;
   232		}
   233		await Promise.all([
   234			flushSignal({
   235				url: r.tracesUrl,
   236				headers: r.headers,
   237				buffer: spans,
   238				body: (items) => makeTracesBody(items, r),
   239				state: tracesState,
   240				signal: "traces",
   241				transport,
   242				logPrefix
   243			}),
   244			flushSignal({
   245				url: r.logsUrl,

--- @maple-dev/effect-sdk/dist/client-BR_-p5Xk.mjs  lines 804-873  (NEW: KEEPALIVE_BUDGET_BYTES shared budget + byteLength + postToIngest (used by session meta/events/blob only)) ---
   804	/**
   805	* Combined budget for this SDK's in-flight `keepalive` bodies.
   806	*
   807	* The Fetch spec caps the *combined* in-flight keepalive body at 64 KiB per
   808	* page, and the browser rejects a request that would cross it. On the way out
   809	* up to three of our writes go at once (metadata row, final events batch, last
   810	* replay chunk), so the budget has to be shared rather than checked per
   811	* request. It stops short of 64 KiB because the OTLP trace exporter spends
   812	* from the same page-wide allowance under its own accounting.
   813	*
   814	* Over the budget a write goes out as a normal request, which the page may or
   815	* may not survive long enough to finish: strictly better than a guaranteed
   816	* rejection.
   817	*/
   818	const KEEPALIVE_BUDGET_BYTES = 49152;
   819	/** On `globalThis`: two bundled SDK copies still share one page-wide allowance. */
   820	const KEEPALIVE_KEY = "__MAPLE_KEEPALIVE_INFLIGHT__";
   821	function keepaliveInflight() {
   822		const owner = globalThis;
   823		const existing = owner[KEEPALIVE_KEY];
   824		if (existing) return existing;
   825		const fresh = { bytes: 0 };
   826		owner[KEEPALIVE_KEY] = fresh;
   827		return fresh;
   828	}
   829	/**
   830	* Reserve `bytes` of the shared keepalive budget. Returns the release function
   831	* when the request may use `keepalive`, `undefined` when it must not.
   832	*/
   833	function reserveKeepalive(requested, bytes) {
   834		if (!requested) return void 0;
   835		const inflight = keepaliveInflight();
   836		if (inflight.bytes + bytes > KEEPALIVE_BUDGET_BYTES) return void 0;
   837		inflight.bytes += bytes;
   838		let released = false;
   839		return () => {
   840			if (released) return;
   841			released = true;
   842			inflight.bytes = Math.max(0, inflight.bytes - bytes);
   843		};
   844	}
   845	/** Body size in bytes: the keepalive limit counts bytes, and `string.length` counts UTF-16 units. */
   846	function byteLength(body) {
   847		return typeof body === "string" ? new TextEncoder().encode(body).byteLength : body.byteLength;
   848	}
   849	/**
   850	* POST to ingest, spending the shared keepalive budget when `keepalive` is
   851	* requested. Resolves with the status only: the response body is cancelled
   852	* before the reservation is released, because the browser counts a keepalive
   853	* request against the page-wide limit until its response body ends, not
   854	* until headers arrive. Rejects exactly as `fetch` does; callers own the
   855	* error policy.
   856	*/
   857	async function postToIngest(url, headers, body, keepalive) {
   858		const release = reserveKeepalive(keepalive, byteLength(body));
   859		try {
   860			const response = await fetch(url, {
   861				method: "POST",
   862				headers,
   863				body,
   864				keepalive: release !== void 0
   865			});
   866			await response.body?.cancel().catch(() => {});
   867			return {
   868				ok: response.ok,
   869				status: response.status
   870			};
   871		} finally {
   872			release?.();
   873		}

--- @maple-dev/effect-sdk/dist/client-BR_-p5Xk.mjs  lines 2446-2458  (MapleFlush keepaliveTransport - unchanged: always keepalive:true, no budget reservation, no size check) ---
  2446	/** Default auto-flush cadence (ms), matching `Otlp.layerJson`'s 5s export interval. */
  2447	const DEFAULT_AUTO_FLUSH_MS = 5e3;
  2448	const browserInstanceId = globalThis.crypto?.randomUUID?.() ?? `browser-${Date.now().toString(36)}-${Math.random().toString(36).slice(2)}`;
  2449	/** `fetch(keepalive)` transport — see file header for why not `sendBeacon`. */
  2450	const keepaliveTransport = { post: async (url, headers, body) => {
  2451		const res = await fetch(url, {
  2452			method: "POST",
  2453			headers,
  2454			body: JSON.stringify(body),
  2455			keepalive: true
  2456		});
  2457		if (!res.ok) throw new Error(`OTLP ${res.status} ${res.statusText}`);
  2458	} };

--- @maple-dev/effect-sdk/dist/replay-BNkcfu5w.mjs  lines 10764-10812  (replay uploadChunk - chunk seq claimed, then async gzip, then POST (unchanged ordering)) ---
 10764	/**
 10765	* gzip and POST one chunk. `seq` is claimed here, monotonic across reloads
 10766	* (persisted on the session record), so a refresh continues the sequence
 10767	* instead of overwriting the previous load's blobs.
 10768	*/
 10769	async function uploadChunk(config, sessionId, body, chunk, keepalive) {
 10770		const chunkSeq = nextChunkSeq();
 10771		let gzipped;
 10772		try {
 10773			gzipped = await gzip(new TextEncoder().encode(body));
 10774		} catch (error) {
 10775			warnDropped("chunk compression", error);
 10776			return;
 10777		}
 10778		return postSessionBlob(config, {
 10779			sessionId,
 10780			chunkSeq,
 10781			...chunk
 10782		}, gzipped, keepalive);
 10783	}
 10784	function startRecording(config, sessionId) {
 10785		let parts = [];
 10786		let bufferBytes = 0;
 10787		let bufferHasCheckpoint = false;
 10788		let firstTimestamp = 0;
 10789		let lastTimestamp = 0;
 10790		let droppedChunk = false;
 10791		let clickCount = 0;
 10792		let stopped = false;
 10793		let exhausted = false;
 10794		const resetBuffer = () => {
 10795			parts = [];
 10796			bufferBytes = 0;
 10797			bufferHasCheckpoint = false;
 10798			firstTimestamp = 0;
 10799			lastTimestamp = 0;
 10800		};
 10801		const flush = async (keepalive = false) => {
 10802			if (stopped || exhausted || parts.length === 0) return;
 10803			const body = `[${parts.join(",")}]`;
 10804			const isCheckpoint = bufferHasCheckpoint;
 10805			const eventCount = parts.length;
 10806			const durationMs = Math.max(0, lastTimestamp - firstTimestamp);
 10807			resetBuffer();
 10808			if (await uploadChunk(config, sessionId, body, {
 10809				isCheckpoint,
 10810				eventCount,
 10811				durationMs
 10812			}, keepalive) === "exhausted" && !exhausted) {

@@@@@@@@@@ END FILE: sdk-excerpts/12-v0.10.0-what-changed-and-what-did-not.txt

@@@@@@@@@@ FILE: evidence/headless-new/runner-console.txt (29525 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112",
  startedAt: "2026-10-07T23:50:02.862Z",
  headlessMode: "new headless (full Chromium, channel 'chromium')",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.0.0 Safari/537.36",
  groups: [ "B", "G", "K", "E" ],
}
[B-probe/probe#0] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-probe/probe#1] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-sdk-concurrent-signals/manual-flush#0] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9
[B-sdk-concurrent-signals/manual-flush#1] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9
[B-sdk-concurrent-signals/reload#0] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9
[B-sdk-concurrent-signals/reload#1] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4183/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4183/dist/app.js:12655:21)
    at try (http://localhost:4183/dist/app.js:8521:26)
    at http://localhost:4183/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4183/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4183/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4183/dist/app.js:2741:23)
    at http://localhost:4183/dist/app.js:3827:9
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#0] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#1] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#0] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#1] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
[G-playwright-keepalive-artifact/steady-state-no-navigation#0] no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/steady-state-no-navigation#1] no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/with-final-reload#0] ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/with-final-reload#1] ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[K-reload-150ms-after-action/replay-on(default)#0] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1856; requests issued around the reload (Playwright): traces(1162B)@1871 logs(1011B)@1871 sessionReplays/meta(1160B)@1872 sessionEvents(1019B)@1872 sessionReplays/meta(1056B)@1931; Playwright failed: sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1865 logs@1867 sessionReplays/meta@1867 sessionEvents@1868 sessionReplays/meta@1930; console: [MapleClientSDK] logs flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1873 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1873 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1872 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1872
[K-reload-150ms-after-action/replay-on(default)#1] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1880; requests issued around the reload (Playwright): traces(1162B)@1899 logs(1011B)@1899 sessionReplays/meta(1160B)@1900 sessionEvents(1019B)@1900 sessionReplays/meta(1056B)@1964; Playwright failed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: traces@1890 logs@1891 sessionReplays/meta@1893 sessionEvents@1895 sessionReplays/meta@1963; console: [MapleClientSDK] traces flush failed; cooldown 60s@1901 | [MapleClientSDK] logs flush failed; cooldown 60s@1901 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1901 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1902 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1900 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1900
[K-reload-150ms-after-action/replay-off#0] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1833; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1846 sessionEvents(498B)@1847 traces(1162B)@1847 logs(1011B)@1847 sessionReplays/meta(1066B)@1888; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1842 traces@1844 sessionEvents@1844 logs@1845 sessionReplays/meta@1886; console: [MapleClientSDK] traces flush failed; cooldown 60s@1852 | [MapleClientSDK] logs flush failed; cooldown 60s@1852 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1853 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1853 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1847
[K-reload-150ms-after-action/replay-off#1] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1819; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1831 sessionEvents(498B)@1831 traces(1162B)@1831 logs(1011B)@1831 sessionReplays/meta(1066B)@1868; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1827 sessionEvents@1828 traces@1828 logs@1830 sessionReplays/meta@1867; console: [MapleClientSDK] traces flush failed; cooldown 60s@1832 | [MapleClientSDK] logs flush failed; cooldown 60s@1832 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1832 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1832 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1831

Scenario summary (new headless (full Chromium, channel 'chromium'), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:50:02.862Z, finished 2026-10-07T23:53:57.374Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/headless-new/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/headless-new/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run1.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/reload-run1.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json

G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
  run 0: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
  run 1: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json

G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
  run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run0.json
  run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 1/2; span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 1/2]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1856; requests issued around the reload (Playwright): traces(1162B)@1871 logs(1011B)@1871 sessionReplays/meta(1160B)@1872 sessionEvents(1019B)@1872 sessionReplays/meta(1056B)@1931; Playwright failed: sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1865 logs@1867 sessionReplays/meta@1867 sessionEvents@1868 sessionReplays/meta@1930; console: [MapleClientSDK] logs flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1873 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1873 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1872 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1872
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1880; requests issued around the reload (Playwright): traces(1162B)@1899 logs(1011B)@1899 sessionReplays/meta(1160B)@1900 sessionEvents(1019B)@1900 sessionReplays/meta(1056B)@1964; Playwright failed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: traces@1890 logs@1891 sessionReplays/meta@1893 sessionEvents@1895 sessionReplays/meta@1963; console: [MapleClientSDK] traces flush failed; cooldown 60s@1901 | [MapleClientSDK] logs flush failed; cooldown 60s@1901 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1901 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1902 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1900 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1900
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run1.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1833; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1846 sessionEvents(498B)@1847 traces(1162B)@1847 logs(1011B)@1847 sessionReplays/meta(1066B)@1888; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1842 traces@1844 sessionEvents@1844 logs@1845 sessionReplays/meta@1886; console: [MapleClientSDK] traces flush failed; cooldown 60s@1852 | [MapleClientSDK] logs flush failed; cooldown 60s@1852 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1853 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1853 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1847
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1819; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1831 sessionEvents(498B)@1831 traces(1162B)@1831 logs(1011B)@1831 sessionReplays/meta(1066B)@1868; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1827 sessionEvents@1828 traces@1828 logs@1830 sessionReplays/meta@1867; console: [MapleClientSDK] traces flush failed; cooldown 60s@1832 | [MapleClientSDK] logs flush failed; cooldown 60s@1832 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1832 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1832 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1831
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-off-run1.json


summary written to ../evidence/headless-new/summary-new-2026-10-07T23-50-02-862Z.txt
@@@@@@@@@@ END FILE: evidence/headless-new/runner-console.txt

@@@@@@@@@@ FILE: evidence/headless-new/summary-new-2026-10-07T23-50-02-862Z.txt (13911 bytes)
Scenario summary (new headless (full Chromium, channel 'chromium'), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:50:02.862Z, finished 2026-10-07T23:53:57.374Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/headless-new/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/headless-new/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/manual-flush-run1.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38469 logs:32132 metrics:928]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/headless-new/B-sdk-concurrent-signals/reload-run1.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18704] meta-ended[2698] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18704B) sessionReplays/meta(2698B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60649] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60649B) sessionReplays/meta(6373B) sessionEvents(21011B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6382] events[496] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6382B) sessionEvents(496B) traces(60649B) sessionReplays/meta(1064B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1124] events[2097] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1124B) sessionEvents(2097B) sessionReplays/meta(1054B)
      evidence: ../evidence/headless-new/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json

G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
  run 0: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
  run 1: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json

G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
  run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run0.json
  run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/headless-new/G-playwright-keepalive-artifact/with-final-reload-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 1/2; span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 1/2]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1856; requests issued around the reload (Playwright): traces(1162B)@1871 logs(1011B)@1871 sessionReplays/meta(1160B)@1872 sessionEvents(1019B)@1872 sessionReplays/meta(1056B)@1931; Playwright failed: sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1865 logs@1867 sessionReplays/meta@1867 sessionEvents@1868 sessionReplays/meta@1930; console: [MapleClientSDK] logs flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush failed; cooldown 60s@1872 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1873 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1873 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1872 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1872
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1880; requests issued around the reload (Playwright): traces(1162B)@1899 logs(1011B)@1899 sessionReplays/meta(1160B)@1900 sessionEvents(1019B)@1900 sessionReplays/meta(1056B)@1964; Playwright failed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: traces@1890 logs@1891 sessionReplays/meta@1893 sessionEvents@1895 sessionReplays/meta@1963; console: [MapleClientSDK] traces flush failed; cooldown 60s@1901 | [MapleClientSDK] logs flush failed; cooldown 60s@1901 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1901 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1902 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1900 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1900
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-on(default)-run1.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1833; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1846 sessionEvents(498B)@1847 traces(1162B)@1847 logs(1011B)@1847 sessionReplays/meta(1066B)@1888; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1842 traces@1844 sessionEvents@1844 logs@1845 sessionReplays/meta@1886; console: [MapleClientSDK] traces flush failed; cooldown 60s@1852 | [MapleClientSDK] logs flush failed; cooldown 60s@1852 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1853 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1853 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1847
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1819; requests issued around the reload (Playwright): sessionReplays/meta(1170B)@1831 sessionEvents(498B)@1831 traces(1162B)@1831 logs(1011B)@1831 sessionReplays/meta(1066B)@1868; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1827 sessionEvents@1828 traces@1828 logs@1830 sessionReplays/meta@1867; console: [MapleClientSDK] traces flush failed; cooldown 60s@1832 | [MapleClientSDK] logs flush failed; cooldown 60s@1832 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1832 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1832 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1831
      evidence: ../evidence/headless-new/K-reload-150ms-after-action/replay-off-run1.json

@@@@@@@@@@ END FILE: evidence/headless-new/summary-new-2026-10-07T23-50-02-862Z.txt

@@@@@@@@@@ FILE: evidence/runner-console-A.txt (5750 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112",
  startedAt: "2026-10-07T23:39:09.004Z",
  headlessMode: "chromium-headless-shell (Playwright default)",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36",
  groups: [ "A" ],
}
[A-oversized-batch/steady30spansPerSec-10s-interval5000(default)#1] NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
[A-oversized-batch/steady40spansPerSec-10s-interval5000(default)#1] PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
[A-oversized-batch/steady40spansPerSec-10s-interval5000(default)#0] PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
[A-oversized-batch/steady30spansPerSec-10s-interval5000(default)#0] NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
[A-oversized-batch/burst200x500B-interval1000#1] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2565,63567,124567ms
[A-oversized-batch/burst200x500B-interval1000#0] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2548,63547,124547ms
[A-oversized-batch/burst200x500B-interval5000(default)#0] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5491,70488,135493ms
[A-oversized-batch/burst200x500B-interval5000(default)#1] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5483,70483,135481ms

Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:39:09.004Z, finished 2026-10-07T23:41:28.743Z

A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
  run 1: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
  run 0: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
  run 1: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: ../evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
  run 0: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: ../evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2565,63567,124567ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval1000-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2548,63547,124547ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval1000-run0.json

A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5491,70488,135493ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval5000(default)-run0.json
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5483,70483,135481ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval5000(default)-run1.json


summary written to ../evidence/summary-shell-2026-10-07T23-39-09-004Z.txt
@@@@@@@@@@ END FILE: evidence/runner-console-A.txt

@@@@@@@@@@ FILE: evidence/runner-console-F-rerun.txt (900 bytes)
[F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight#0] REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
[F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight#1] REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
[F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight#0] REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
[F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight#1] REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
@@@@@@@@@@ END FILE: evidence/runner-console-F-rerun.txt

@@@@@@@@@@ FILE: evidence/runner-console-W.txt (2528 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112",
  startedAt: "2026-10-07T23:58:29.444Z",
  headlessMode: "chromium-headless-shell (Playwright default)",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36",
  groups: [ "W" ],
}
[W-workaround-keepalive-downgrade-shim/without-shim#0] LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
[W-workaround-keepalive-downgrade-shim/without-shim#1] LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
[W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim#0] all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
[W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim#1] all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0

Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:58:29.444Z, finished 2026-10-07T23:59:04.742Z

W-workaround-keepalive-downgrade-shim / without-shim  [LOSS (burst 0/200, ticks 0/5): 2/2]
  run 0: LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/without-shim-run0.json
  run 1: LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/without-shim-run1.json

W-workaround-keepalive-downgrade-shim / with-keepalive-downgrade-shim  [all delivered: 2/2]
  run 0: all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run0.json
  run 1: all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run1.json


summary written to ../evidence/summary-shell-2026-10-07T23-58-29-444Z.txt
@@@@@@@@@@ END FILE: evidence/runner-console-W.txt

@@@@@@@@@@ FILE: evidence/runner-console-main.txt (102949 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.8.2 + effect@4.0.0-rc.112",
  startedAt: "2026-10-07T23:39:09.001Z",
  headlessMode: "chromium-headless-shell (Playwright default)",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36",
  groups: [
    "B", "A0", "C", "D", "Dr", "E", "F", "G", "H", "I", "K", "L"
  ],
}
[B-probe/probe#0] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-probe/probe#1] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-sdk-concurrent-signals/manual-flush#0] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[B-sdk-concurrent-signals/manual-flush#1] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[B-sdk-concurrent-signals/manual-flush#2] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[B-sdk-concurrent-signals/reload#0] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[B-sdk-concurrent-signals/reload#1] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[B-sdk-concurrent-signals/reload#2] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9 | [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[A0-threshold/n100#0] DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
[A0-threshold/n150#0] DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
[A0-threshold/n160#0] DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
[A0-threshold/n163#0] DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
[A0-threshold/n164#0] REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[A0-threshold/n165#0] REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[A0-threshold/n170#0] REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[A0-threshold/n200#0] REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at Object.post (http://localhost:4173/dist/app.js:12655:21)
    at try (http://localhost:4173/dist/app.js:8521:26)
    at http://localhost:4173/dist/app.js:2990:26
    at ~effect/Utils/internal (http://localhost:4173/dist/app.js:833:14)
    at MixedScheduler.<anonymous> (http://localhost:4173/dist/app.js:2990:7)
    at ~effect/Effect/evaluate (http://localhost:4173/dist/app.js:3005:22)
    at FiberImpl.runLoop (http://localhost:4173/dist/app.js:2779:107)
    at FiberImpl.evaluate (http://localhost:4173/dist/app.js:2741:23)
    at http://localhost:4173/dist/app.js:3827:9
[C-unload-during-cooldown/inject503-then-reload#0] REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2937,t=2942
[C-unload-during-cooldown/inject503-then-reload#1] REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2947,t=2956
[C-unload-during-cooldown/inject503-then-reload#2] REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2122,t=2952,t=2958
[C-unload-during-cooldown/inject503-then-navigate#0] REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2937,t=2941
[C-unload-during-cooldown/inject503-then-navigate#1] REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2108,t=2937,t=2943
[C-unload-during-cooldown/inject503-then-navigate#2] REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2147,t=2960,t=2965
[C-unload-during-cooldown/inject503-then-close#0] REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2938,t=2942
[C-unload-during-cooldown/inject503-then-close#1] REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2930,t=2935
[C-unload-during-cooldown/inject503-then-close#2] REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2939,t=2945
[C-unload-during-cooldown/control-no-failure-reload#0] CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2972
[C-unload-during-cooldown/control-no-failure-reload#1] CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2949
[C-unload-during-cooldown/control-no-failure-reload#2] CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2970
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000#0] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5125ms); D-tail delivered=false; reload at t=1401ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000#1] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5107ms); D-tail delivered=false; reload at t=1382ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000#2] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5108ms); D-tail delivered=false; reload at t=1375ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000#0] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9124ms); D-tail delivered=false; reload at t=5383ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000#1] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9115ms); D-tail delivered=false; reload at t=5367ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000#2] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9114ms); D-tail delivered=false; reload at t=5373ms
[D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload#0] REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
[D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload#1] REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
[D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload#2] REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms
[D-pagehide-behind-inflight-flush/control-no-latency-interval1000#0] CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1405ms); reload at t=1393ms
[D-pagehide-behind-inflight-flush/control-no-latency-interval1000#1] CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1412ms); reload at t=1406ms
[D-pagehide-behind-inflight-flush/control-no-latency-interval1000#2] CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1123ms); D-tail delivered=true (received t=1401ms); reload at t=1394ms
[D-hang-head-of-line-blocking/first-POST-held-15s#0] REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1153ms): span export delay max 14990 ms / min 909 ms; log export delay max 14991 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17143:1sp/1157B 18143:1sp/1157B 19144:1sp/1157B]
[D-hang-head-of-line-blocking/first-POST-held-15s#1] REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1137ms): span export delay max 14977 ms / min 905 ms; log export delay max 14977 ms (logs endpoint itself is fast); traces POSTs at server t=[1116:1sp/1160B 16120:15sp/6677B 17115:1sp/1157B 18115:1sp/1157B 19115:1sp/1157B]
[Dr-random-reload-latency/latency150-interval1000#0] no loss — latency 150 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#1] no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#2] no loss — latency 150 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#3] no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#4] no loss — latency 150 ms, interval 1000: created 38, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#5] no loss — latency 150 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#6] no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#7] LOSS (1 spans) — latency 150 ms, interval 1000: created 40, lost 1 (created 100 ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency150-interval1000#8] no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency150-interval1000#9] no loss — latency 150 ms, interval 1000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval1000#0] no loss — latency 600 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval1000#1] no loss — latency 600 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval1000#2] LOSS (5 spans) — latency 600 ms, interval 1000: created 54, lost 5 (created 513,411,305,202,100 ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval1000#3] LOSS (2 spans) — latency 600 ms, interval 1000: created 22, lost 2 (created 202,100 ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval1000#4] no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval1000#5] no loss — latency 600 ms, interval 1000: created 57, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval1000#6] no loss — latency 600 ms, interval 1000: created 39, lost 0 (created  ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval1000#7] LOSS (4 spans) — latency 600 ms, interval 1000: created 34, lost 4 (created 406,304,202,100 ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval1000#8] no loss — latency 600 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval1000#9] no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#0] no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#1] no loss — latency 600 ms, interval 5000: created 41, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#2] no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval5000#3] no loss — latency 600 ms, interval 5000: created 22, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#4] no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
[Dr-random-reload-latency/latency600-interval5000#5] no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#6] no loss — latency 600 ms, interval 5000: created 35, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#7] no loss — latency 600 ms, interval 5000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#8] no loss — latency 600 ms, interval 5000: created 21, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/latency600-interval5000#9] no loss — latency 600 ms, interval 5000: created 25, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#0] no loss — latency 0 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#1] no loss — latency 0 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#2] no loss — latency 0 ms, interval 1000: created 43, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#3] no loss — latency 0 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#4] no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#5] no loss — latency 0 ms, interval 1000: created 31, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#6] no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#7] no loss — latency 0 ms, interval 1000: created 33, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#8] no loss — latency 0 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
[Dr-random-reload-latency/control-latency0-interval1000#9] no loss — latency 0 ms, interval 1000: created 23, lost 0 (created  ms before reload); trace flush in flight at reload: false
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#2] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#0] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#1] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#2] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#0] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#1] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#2] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#2] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#2] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
[F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight#0] REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at flush (http://localhost:4173/dist/app.js:11993:11)
    at http://localhost:4173/dist/app.js:12025:45; Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight#1] REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at flush (http://localhost:4173/dist/app.js:11993:11)
    at http://localhost:4173/dist/app.js:12025:45; Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight#0] REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at startRun (http://localhost:4173/dist/app.js:12288:5)
    at startSessionLifecycle (http://localhost:4173/dist/app.js:12345:3)
    at startReplaySession (http://localhost:4173/dist/chunk-9ehzfcba.js:12322:10)
    at http://localhost:4173/dist/app.js:12573:23; Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight#1] REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at startRun (http://localhost:4173/dist/app.js:12288:5)
    at startSessionLifecycle (http://localhost:4173/dist/app.js:12345:3)
    at startReplaySession (http://localhost:4173/dist/chunk-9ehzfcba.js:12322:10)
    at http://localhost:4173/dist/app.js:12573:23; Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]
[Fr-random-reloads/flush-8-random-reloads-latency300#0] REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"traces[net::ERR_ABORTED]":1,"sessionReplays/meta[net::ERR_ABORTED]":8}
[Fr-random-reloads/flush-8-random-reloads-latency300#1] REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST; [maple] session replay blob PUT], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":5,"sessionReplays/blob[net::ERR_ABORTED]":1,"traces[net::ERR_ABORTED]":1}
[F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight#0] REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33); Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight#1] REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at endRun (http://localhost:4173/dist/app.js:12308:7)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33); Playwright requestfailed: sessionReplays/meta[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight#0] REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at startRun (http://localhost:4173/dist/app.js:12288:5)
    at startSessionLifecycle (http://localhost:4173/dist/app.js:12345:3)
    at startReplaySession (http://localhost:4173/dist/chunk-9ehzfcba.js:12322:10)
    at http://localhost:4173/dist/app.js:12573:23; Playwright requestfailed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]
[F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight#1] REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionEvents (http://localhost:4173/dist/app.js:11582:9)
    at Object.flush (http://localhost:4173/dist/app.js:11993:11)
    at Object.onSuspend (http://localhost:4173/dist/chunk-9ehzfcba.js:12341:96)
    at endRun (http://localhost:4173/dist/app.js:12309:29)
    at onPageHide (http://localhost:4173/dist/app.js:12322:33) | [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
    at window.fetch (http://localhost:4173/dist/chunk-9ehzfcba.js:71:27)
    at postSessionMeta (http://localhost:4173/dist/app.js:11564:9)
    at Object.post (http://localhost:4173/dist/chunk-9ehzfcba.js:12328:7)
    at post (http://localhost:4173/dist/app.js:12253:11)
    at startRun (http://localhost:4173/dist/app.js:12288:5)
    at startSessionLifecycle (http://localhost:4173/dist/app.js:12345:3)
    at startReplaySession (http://localhost:4173/dist/chunk-9ehzfcba.js:12322:10)
    at http://localhost:4173/dist/app.js:12573:23; Playwright requestfailed: sessionEvents[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]
[Fr-random-reloads/layer-8-random-reloads-latency300#0] REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"sessionReplays/meta[net::ERR_ABORTED]":8}
[Fr-random-reloads/layer-8-random-reloads-latency300#1] REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":6}
[G-playwright-keepalive-artifact/steady-state-no-navigation#0] no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/steady-state-no-navigation#1] no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/with-final-reload#0] ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[G-playwright-keepalive-artifact/with-final-reload#1] ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
[H-maple-layer-unload/layer#0] tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
[H-maple-layer-unload/layer#1] tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
[H-maple-layer-unload/layer#2] tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
[H-maple-layer-unload/layer-flusher#0] tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[H-maple-layer-unload/layer-flusher#1] tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[H-maple-layer-unload/layer-flusher#2] tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[H-maple-layer-unload/flush#0] tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[H-maple-layer-unload/flush#1] tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[H-maple-layer-unload/flush#2] tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
[I-cross-origin/same-origin#0] tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
[I-cross-origin/same-origin#1] tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
[I-cross-origin/cross-origin#0] tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
[I-cross-origin/cross-origin#1] tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
[K-reload-150ms-after-action/replay-on(default)#0] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1795; requests issued around the reload (Playwright): traces(1165B)@1807 logs(1014B)@1807 sessionReplays/meta(1164B)@1807 sessionEvents(1021B)@1807 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1801 logs@1802 sessionReplays/meta@1802 sessionEvents@1804 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1808 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1808 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1808 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1808
[K-reload-150ms-after-action/replay-on(default)#1] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1810; requests issued around the reload (Playwright): traces(1165B)@1824 logs(1014B)@1824 sessionReplays/meta(1164B)@1824 sessionEvents(1021B)@1824 sessionReplays/meta(1060B)@1891; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1817 sessionReplays/meta@1818 logs@1818 sessionEvents@1820 sessionReplays/meta@1890; console: [MapleClientSDK] logs flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1825 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1825 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1825 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1825
[K-reload-150ms-after-action/replay-on(default)#2] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1803; requests issued around the reload (Playwright): traces(1165B)@1815 logs(1014B)@1816 sessionReplays/meta(1164B)@1816 sessionEvents(1021B)@1816 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1809 logs@1810 sessionReplays/meta@1813 sessionEvents@1814 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1816 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1816 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1816 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1816
[K-reload-150ms-after-action/replay-off#0] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1770; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1781 sessionEvents(499B)@1781 traces(1165B)@1782 logs(1014B)@1782 sessionReplays/meta(1070B)@1814; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1777 traces@1778 sessionEvents@1779 logs@1781 sessionReplays/meta@1811; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1787 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1787 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1788 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1782
[K-reload-150ms-after-action/replay-off#1] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1771; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1783 sessionEvents(499B)@1783 traces(1165B)@1784 logs(1014B)@1784 sessionReplays/meta(1070B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1778 sessionEvents@1779 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1785 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1785 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1785 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
[K-reload-150ms-after-action/replay-off#2] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1769; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1780 sessionEvents(499B)@1780 traces(1165B)@1781 logs(1014B)@1781 sessionReplays/meta(1070B)@1813; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1776 traces@1777 sessionEvents@1778 logs@1780 sessionReplays/meta@1810; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1782 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1782 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1782 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1781
[L-multi-arg-log-body/logError-4-args#0] body is an OTLP arrayValue (structured), not a JSON string — OTLP logRecord.body as POSTed by the SDK: {"arrayValue":{"values":[{"stringValue":"L Failed to connect"},{"stringValue":"Error: ClientError: boom"},{"stringValue":"Unknown Error"},{"stringValue":"{\"code\":42,\"nested\":{\"a\":1}}"}]}}

Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:39:09.001Z, finished 2026-10-07T23:58:20.324Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 3/3]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run1.json
  run 2: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run2.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 3/3]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run1.json
  run 2: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run2.json

A0-threshold / n100  [DELIVERED: 1/1]
  run 0: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/A0-threshold/n100-run0.json

A0-threshold / n150  [DELIVERED: 1/1]
  run 0: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/A0-threshold/n150-run0.json

A0-threshold / n160  [DELIVERED: 1/1]
  run 0: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/A0-threshold/n160-run0.json

A0-threshold / n163  [DELIVERED: 1/1]
  run 0: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/A0-threshold/n163-run0.json

A0-threshold / n164  [REJECTED: 1/1]
  run 0: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n164-run0.json

A0-threshold / n165  [REJECTED: 1/1]
  run 0: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n165-run0.json

A0-threshold / n170  [REJECTED: 1/1]
  run 0: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n170-run0.json

A0-threshold / n200  [REJECTED: 1/1]
  run 0: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n200-run0.json

C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2937,t=2942
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2947,t=2956
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2122,t=2952,t=2958
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run2.json

C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2937,t=2941
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2108,t=2937,t=2943
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2147,t=2960,t=2965
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run2.json

C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2938,t=2942
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2930,t=2935
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2939,t=2945
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run2.json

C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 3/3]
  run 0: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2972
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run0.json
  run 1: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2949
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run1.json
  run 2: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2970
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run2.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5125ms); D-tail delivered=false; reload at t=1401ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5107ms); D-tail delivered=false; reload at t=1382ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5108ms); D-tail delivered=false; reload at t=1375ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run2.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9124ms); D-tail delivered=false; reload at t=5383ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9115ms); D-tail delivered=false; reload at t=5367ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9114ms); D-tail delivered=false; reload at t=5373ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run2.json

D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run2.json

D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 3/3]
  run 0: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1405ms); reload at t=1393ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
  run 1: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1412ms); reload at t=1406ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json
  run 2: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1123ms); D-tail delivered=true (received t=1401ms); reload at t=1394ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run2.json

D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
  run 0: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1153ms): span export delay max 14990 ms / min 909 ms; log export delay max 14991 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17143:1sp/1157B 18143:1sp/1157B 19144:1sp/1157B]
      evidence: ../evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
  run 1: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1137ms): span export delay max 14977 ms / min 905 ms; log export delay max 14977 ms (logs endpoint itself is fast); traces POSTs at server t=[1116:1sp/1160B 16120:15sp/6677B 17115:1sp/1157B 18115:1sp/1157B 19115:1sp/1157B]
      evidence: ../evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json

Dr-random-reload-latency / latency150-interval1000  [no loss: 9/10; LOSS (1 spans): 1/10]
  run 0: no loss — latency 150 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run0.json
  run 1: no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run1.json
  run 2: no loss — latency 150 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run2.json
  run 3: no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run3.json
  run 4: no loss — latency 150 ms, interval 1000: created 38, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run4.json
  run 5: no loss — latency 150 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run5.json
  run 6: no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run6.json
  run 7: LOSS (1 spans) — latency 150 ms, interval 1000: created 40, lost 1 (created 100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run7.json
  run 8: no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run8.json
  run 9: no loss — latency 150 ms, interval 1000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run9.json

Dr-random-reload-latency / latency600-interval1000  [no loss: 7/10; LOSS (5 spans): 1/10; LOSS (2 spans): 1/10; LOSS (4 spans): 1/10]
  run 0: no loss — latency 600 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run0.json
  run 1: no loss — latency 600 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run1.json
  run 2: LOSS (5 spans) — latency 600 ms, interval 1000: created 54, lost 5 (created 513,411,305,202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run2.json
  run 3: LOSS (2 spans) — latency 600 ms, interval 1000: created 22, lost 2 (created 202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run3.json
  run 4: no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run4.json
  run 5: no loss — latency 600 ms, interval 1000: created 57, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run5.json
  run 6: no loss — latency 600 ms, interval 1000: created 39, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run6.json
  run 7: LOSS (4 spans) — latency 600 ms, interval 1000: created 34, lost 4 (created 406,304,202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run7.json
  run 8: no loss — latency 600 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run8.json
  run 9: no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run9.json

Dr-random-reload-latency / latency600-interval5000  [no loss: 10/10]
  run 0: no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run0.json
  run 1: no loss — latency 600 ms, interval 5000: created 41, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run1.json
  run 2: no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run2.json
  run 3: no loss — latency 600 ms, interval 5000: created 22, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run3.json
  run 4: no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run4.json
  run 5: no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run5.json
  run 6: no loss — latency 600 ms, interval 5000: created 35, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run6.json
  run 7: no loss — latency 600 ms, interval 5000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run7.json
  run 8: no loss — latency 600 ms, interval 5000: created 21, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run8.json
  run 9: no loss — latency 600 ms, interval 5000: created 25, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run9.json

Dr-random-reload-latency / control-latency0-interval1000  [no loss: 10/10]
  run 0: no loss — latency 0 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run0.json
  run 1: no loss — latency 0 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run1.json
  run 2: no loss — latency 0 ms, interval 1000: created 43, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run2.json
  run 3: no loss — latency 0 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run3.json
  run 4: no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run4.json
  run 5: no loss — latency 0 ms, interval 1000: created 31, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run5.json
  run 6: no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run6.json
  run 7: no loss — latency 0 ms, interval 1000: created 33, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run7.json
  run 8: no loss — latency 0 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run8.json
  run 9: no loss — latency 0 ms, interval 1000: created 23, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run9.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run2.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 3/3]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json
  run 2: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run2.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 3/3]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json
  run 2: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run2.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run2.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run2.json

F-replay-posts-aborted-by-navigation / flush-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run0.json
  run 1: REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run1.json

F-replay-posts-aborted-by-navigation / flush-active-meta-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run0.json
  run 1: REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run1.json

Fr-random-reloads / flush-8-random-reloads-latency300  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"traces[net::ERR_ABORTED]":1,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: ../evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run0.json
  run 1: REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST; [maple] session replay blob PUT], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":5,"sessionReplays/blob[net::ERR_ABORTED]":1,"traces[net::ERR_ABORTED]":1}
      evidence: ../evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run1.json

F-replay-posts-aborted-by-navigation / layer-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run0.json
  run 1: REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run1.json

F-replay-posts-aborted-by-navigation / layer-active-meta-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run0.json
  run 1: REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run1.json

Fr-random-reloads / layer-8-random-reloads-latency300  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: ../evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run0.json
  run 1: REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":6}
      evidence: ../evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run1.json

G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
  run 0: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
  run 1: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json

G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
  run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/with-final-reload-run0.json
  run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/with-final-reload-run1.json

H-maple-layer-unload / layer  [tail LOST: 3/3]
  run 0: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run0.json
  run 1: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run1.json
  run 2: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run2.json

H-maple-layer-unload / layer-flusher  [tail delivered: 3/3]
  run 0: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run0.json
  run 1: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run1.json
  run 2: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run2.json

H-maple-layer-unload / flush  [tail delivered: 3/3]
  run 0: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run0.json
  run 1: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run1.json
  run 2: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run2.json

I-cross-origin / same-origin  [tail delivered: 2/2]
  run 0: tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/same-origin-run0.json
  run 1: tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/same-origin-run1.json

I-cross-origin / cross-origin  [tail delivered: 2/2]
  run 0: tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/cross-origin-run0.json
  run 1: tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/cross-origin-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 3/3]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1795; requests issued around the reload (Playwright): traces(1165B)@1807 logs(1014B)@1807 sessionReplays/meta(1164B)@1807 sessionEvents(1021B)@1807 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1801 logs@1802 sessionReplays/meta@1802 sessionEvents@1804 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1808 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1808 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1808 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1808
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1810; requests issued around the reload (Playwright): traces(1165B)@1824 logs(1014B)@1824 sessionReplays/meta(1164B)@1824 sessionEvents(1021B)@1824 sessionReplays/meta(1060B)@1891; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1817 sessionReplays/meta@1818 logs@1818 sessionEvents@1820 sessionReplays/meta@1890; console: [MapleClientSDK] logs flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1825 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1825 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1825 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1825
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run1.json
  run 2: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1803; requests issued around the reload (Playwright): traces(1165B)@1815 logs(1014B)@1816 sessionReplays/meta(1164B)@1816 sessionEvents(1021B)@1816 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1809 logs@1810 sessionReplays/meta@1813 sessionEvents@1814 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1816 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1816 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1816 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1816
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run2.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 3/3]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1770; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1781 sessionEvents(499B)@1781 traces(1165B)@1782 logs(1014B)@1782 sessionReplays/meta(1070B)@1814; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1777 traces@1778 sessionEvents@1779 logs@1781 sessionReplays/meta@1811; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1787 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1787 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1788 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1782
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1771; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1783 sessionEvents(499B)@1783 traces(1165B)@1784 logs(1014B)@1784 sessionReplays/meta(1070B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1778 sessionEvents@1779 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1785 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1785 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1785 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run1.json
  run 2: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1769; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1780 sessionEvents(499B)@1780 traces(1165B)@1781 logs(1014B)@1781 sessionReplays/meta(1070B)@1813; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1776 traces@1777 sessionEvents@1778 logs@1780 sessionReplays/meta@1810; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1782 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1782 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1782 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1781
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run2.json

L-multi-arg-log-body / logError-4-args  [body is an OTLP arrayValue (structured), not a JSON string: 1/1]
  run 0: body is an OTLP arrayValue (structured), not a JSON string — OTLP logRecord.body as POSTed by the SDK: {"arrayValue":{"values":[{"stringValue":"L Failed to connect"},{"stringValue":"Error: ClientError: boom"},{"stringValue":"Unknown Error"},{"stringValue":"{\"code\":42,\"nested\":{\"a\":1}}"}]}}
      evidence: ../evidence/L-multi-arg-log-body/logError-4-args-run0.json


summary written to ../evidence/summary-shell-2026-10-07T23-39-09-001Z.txt
@@@@@@@@@@ END FILE: evidence/runner-console-main.txt

@@@@@@@@@@ FILE: evidence/sdk-0.10.0/runner-console-A.txt (5863 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.10.0 (optional variant build)",
  startedAt: "2026-10-07T23:41:29.113Z",
  headlessMode: "chromium-headless-shell (Playwright default)",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36",
  groups: [ "A" ],
}
[A-oversized-batch/steady40spansPerSec-10s-interval5000(default)#1] PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10460,75460ms
[A-oversized-batch/steady30spansPerSec-10s-interval5000(default)#1] NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
[A-oversized-batch/steady30spansPerSec-10s-interval5000(default)#0] NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
[A-oversized-batch/steady40spansPerSec-10s-interval5000(default)#0] PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10481,75482ms
[A-oversized-batch/burst200x500B-interval1000#1] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2447,63447,124447ms
[A-oversized-batch/burst200x500B-interval1000#0] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2513,63514,124513ms
[A-oversized-batch/burst200x500B-interval5000(default)#1] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5451,70449,135456ms
[A-oversized-batch/burst200x500B-interval5000(default)#0] REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5488,70485,135492ms

Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:41:29.113Z, finished 2026-10-07T23:43:48.562Z

A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
  run 1: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10460,75460ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
  run 0: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10481,75482ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
  run 1: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
  run 0: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2447,63447,124447ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2513,63514,124513ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run0.json

A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5451,70449,135456ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5488,70485,135492ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run0.json


summary written to ../evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-41-29-113Z.txt
@@@@@@@@@@ END FILE: evidence/sdk-0.10.0/runner-console-A.txt

@@@@@@@@@@ FILE: evidence/sdk-0.10.0/runner-console-main.txt (49901 bytes)
environment: {
  sdkUnderTest: "@maple-dev/effect-sdk@0.10.0 (optional variant build)",
  startedAt: "2026-10-07T23:43:48.971Z",
  headlessMode: "chromium-headless-shell (Playwright default)",
  chromiumVersion: "143.0.7499.4",
  playwrightVersion: "1.57.0",
  bunVersion: "1.4.2",
  platform: "linux x64",
  userAgent: "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) HeadlessChrome/143.0.7499.4 Safari/537.36",
  groups: [ "A0", "B", "C", "D", "E", "K" ],
}
[B-probe/probe#0] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-probe/probe#1] MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
[B-sdk-concurrent-signals/manual-flush#0] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[B-sdk-concurrent-signals/manual-flush#1] REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[B-sdk-concurrent-signals/reload#0] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[B-sdk-concurrent-signals/reload#1] REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9 | [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n100#0] DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
[A0-threshold/n100#1] DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
[A0-threshold/n150#0] DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
[A0-threshold/n150#1] DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
[A0-threshold/n160#0] DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
[A0-threshold/n160#1] DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
[A0-threshold/n163#0] DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
[A0-threshold/n163#1] DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
[A0-threshold/n164#0] REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n164#1] REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n165#0] REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n165#1] REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n170#0] REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n170#1] REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n200#0] REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[A0-threshold/n200#1] REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
    at http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:47
    at withStartedTraceId (http://localhost:4183/dist-0.10.0/app.js:12140:15)
    at window.fetch (http://localhost:4183/dist-0.10.0/chunk-kz9b2wg9.js:116:22)
    at Object.post (http://localhost:4183/dist-0.10.0/app.js:13602:21)
    at try (http://localhost:4183/dist-0.10.0/app.js:9466:26)
    at MixedScheduler.<anonymous> (http://localhost:4183/dist-0.10.0/app.js:2965:7)
    at ~effect/Effect/evaluate (http://localhost:4183/dist-0.10.0/app.js:2979:38)
    at FiberImpl.runLoop (http://localhost:4183/dist-0.10.0/app.js:2734:95)
    at FiberImpl.evaluate (http://localhost:4183/dist-0.10.0/app.js:2688:23)
    at http://localhost:4183/dist-0.10.0/app.js:3882:9
[C-unload-during-cooldown/inject503-then-reload#0] REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2115,t=2950,t=2956
[C-unload-during-cooldown/inject503-then-reload#1] REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2949,t=2953
[C-unload-during-cooldown/inject503-then-navigate#0] REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2119,t=2954,t=2960
[C-unload-during-cooldown/inject503-then-navigate#1] REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2956,t=2962
[C-unload-during-cooldown/inject503-then-close#0] REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2931,t=2936
[C-unload-during-cooldown/inject503-then-close#1] REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2134,t=2949,t=2959
[C-unload-during-cooldown/control-no-failure-reload#0] CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2960
[C-unload-during-cooldown/control-no-failure-reload#1] CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2954
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000#0] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5114ms); D-tail delivered=false; reload at t=1383ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000#1] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5124ms); D-tail delivered=false; reload at t=1380ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000#0] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9123ms); D-tail delivered=false; reload at t=5383ms
[D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000#1] REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9116ms); D-tail delivered=false; reload at t=5370ms
[D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload#0] REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8113ms); D-tail delivered=false; reload at t=6231ms
[D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload#1] REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6241ms
[D-pagehide-behind-inflight-flush/control-no-latency-interval1000#0] CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1145ms); D-tail delivered=true (received t=1408ms); reload at t=1401ms
[D-pagehide-behind-inflight-flush/control-no-latency-interval1000#1] CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1115ms); D-tail delivered=true (received t=1396ms); reload at t=1390ms
[D-hang-head-of-line-blocking/first-POST-held-15s#0] REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1161ms): span export delay max 14983 ms / min 886 ms; log export delay max 14983 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17144:1sp/1157B 18144:1sp/1157B 19144:1sp/1157B]
[D-hang-head-of-line-blocking/first-POST-held-15s#1] REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1131ms): span export delay max 14992 ms / min 922 ms; log export delay max 14992 ms (logs endpoint itself is fast); traces POSTs at server t=[1123:1sp/1160B 16128:15sp/6677B 17123:1sp/1157B 18123:1sp/1157B 19122:1sp/1157B]
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#0] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines#1] LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#0] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
[E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)#1] LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E4-no-spans-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#0] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
[E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation#1] LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
[K-reload-150ms-after-action/replay-on(default)#0] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1777; requests issued around the reload (Playwright): traces(1165B)@1789 logs(1014B)@1790 sessionReplays/meta(1194B)@1790 sessionEvents(1043B)@1790 sessionReplays/meta(1090B)@1850; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1783 sessionReplays/meta@1785 logs@1785 sessionEvents@1787 sessionReplays/meta@1849; console: [MapleClientSDK] logs flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1790 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1790 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1790 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1790
[K-reload-150ms-after-action/replay-on(default)#1] span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1807; requests issued around the reload (Playwright): traces(1165B)@1819 logs(1014B)@1819 sessionReplays/meta(1194B)@1819 sessionEvents(1043B)@1819 sessionReplays/meta(1090B)@1882; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1813 logs@1814 sessionReplays/meta@1815 sessionEvents@1816 sessionReplays/meta@1882; console: [MapleClientSDK] logs flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1820 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1820 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1819 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1819
[K-reload-150ms-after-action/replay-off#0] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1783; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1792 sessionEvents(510B)@1793 traces(1165B)@1793 logs(1014B)@1793 sessionReplays/meta(1100B)@1825; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1789 sessionEvents@1790 traces@1790 logs@1791 sessionReplays/meta@1822; console: [MapleClientSDK] traces flush failed; cooldown 60s@1793 | [MapleClientSDK] logs flush failed; cooldown 60s@1793 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1793 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1793 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1793
[K-reload-150ms-after-action/replay-off#1] span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1772; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1782 sessionEvents(510B)@1783 traces(1165B)@1783 logs(1014B)@1784 sessionReplays/meta(1100B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1779 sessionEvents@1780 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1784 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1784 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1784 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784

Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:43:48.971Z, finished 2026-10-07T23:50:02.504Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/sdk-0.10.0/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/sdk-0.10.0/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run1.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run1.json

A0-threshold / n100  [DELIVERED: 2/2]
  run 0: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n100-run0.json
  run 1: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n100-run1.json

A0-threshold / n150  [DELIVERED: 2/2]
  run 0: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n150-run0.json
  run 1: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n150-run1.json

A0-threshold / n160  [DELIVERED: 2/2]
  run 0: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n160-run0.json
  run 1: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n160-run1.json

A0-threshold / n163  [DELIVERED: 2/2]
  run 0: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n163-run0.json
  run 1: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n163-run1.json

A0-threshold / n164  [REJECTED: 2/2]
  run 0: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n164-run0.json
  run 1: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n164-run1.json

A0-threshold / n165  [REJECTED: 2/2]
  run 0: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n165-run0.json
  run 1: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n165-run1.json

A0-threshold / n170  [REJECTED: 2/2]
  run 0: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n170-run0.json
  run 1: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n170-run1.json

A0-threshold / n200  [REJECTED: 2/2]
  run 0: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n200-run0.json
  run 1: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n200-run1.json

C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2115,t=2950,t=2956
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2949,t=2953
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run1.json

C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2119,t=2954,t=2960
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2956,t=2962
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run1.json

C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2931,t=2936
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2134,t=2949,t=2959
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run1.json

C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 2/2]
  run 0: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2960
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run0.json
  run 1: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2954
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run1.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5114ms); D-tail delivered=false; reload at t=1383ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5124ms); D-tail delivered=false; reload at t=1380ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9123ms); D-tail delivered=false; reload at t=5383ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9116ms); D-tail delivered=false; reload at t=5370ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json

D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8113ms); D-tail delivered=false; reload at t=6231ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6241ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json

D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 2/2]
  run 0: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1145ms); D-tail delivered=true (received t=1408ms); reload at t=1401ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
  run 1: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1115ms); D-tail delivered=true (received t=1396ms); reload at t=1390ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json

D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
  run 0: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1161ms): span export delay max 14983 ms / min 886 ms; log export delay max 14983 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17144:1sp/1157B 18144:1sp/1157B 19144:1sp/1157B]
      evidence: ../evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
  run 1: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1131ms): span export delay max 14992 ms / min 922 ms; log export delay max 14992 ms (logs endpoint itself is fast); traces POSTs at server t=[1123:1sp/1160B 16128:15sp/6677B 17123:1sp/1157B 18123:1sp/1157B 19122:1sp/1157B]
      evidence: ../evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1777; requests issued around the reload (Playwright): traces(1165B)@1789 logs(1014B)@1790 sessionReplays/meta(1194B)@1790 sessionEvents(1043B)@1790 sessionReplays/meta(1090B)@1850; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1783 sessionReplays/meta@1785 logs@1785 sessionEvents@1787 sessionReplays/meta@1849; console: [MapleClientSDK] logs flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1790 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1790 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1790 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1790
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1807; requests issued around the reload (Playwright): traces(1165B)@1819 logs(1014B)@1819 sessionReplays/meta(1194B)@1819 sessionEvents(1043B)@1819 sessionReplays/meta(1090B)@1882; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1813 logs@1814 sessionReplays/meta@1815 sessionEvents@1816 sessionReplays/meta@1882; console: [MapleClientSDK] logs flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1820 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1820 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1819 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1819
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run1.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1783; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1792 sessionEvents(510B)@1793 traces(1165B)@1793 logs(1014B)@1793 sessionReplays/meta(1100B)@1825; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1789 sessionEvents@1790 traces@1790 logs@1791 sessionReplays/meta@1822; console: [MapleClientSDK] traces flush failed; cooldown 60s@1793 | [MapleClientSDK] logs flush failed; cooldown 60s@1793 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1793 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1793 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1793
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1772; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1782 sessionEvents(510B)@1783 traces(1165B)@1783 logs(1014B)@1784 sessionReplays/meta(1100B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1779 sessionEvents@1780 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1784 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1784 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1784 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run1.json


summary written to ../evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-43-48-971Z.txt
@@@@@@@@@@ END FILE: evidence/sdk-0.10.0/runner-console-main.txt

@@@@@@@@@@ FILE: evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-41-29-113Z.txt (3149 bytes)
Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:41:29.113Z, finished 2026-10-07T23:43:48.562Z

A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
  run 1: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10460,75460ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
  run 0: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 156 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10481,75482ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
  run 1: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
  run 0: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2447,63447,124447ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2513,63514,124513ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval1000-run0.json

A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5451,70449,135456ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5488,70485,135492ms
      evidence: ../evidence/sdk-0.10.0/A-oversized-batch/burst200x500B-interval5000(default)-run0.json

@@@@@@@@@@ END FILE: evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-41-29-113Z.txt

@@@@@@@@@@ FILE: evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-43-48-971Z.txt (22139 bytes)
Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:43:48.971Z, finished 2026-10-07T23:50:02.504Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/sdk-0.10.0/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/sdk-0.10.0/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/manual-flush-run1.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 2/2]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/B-sdk-concurrent-signals/reload-run1.json

A0-threshold / n100  [DELIVERED: 2/2]
  run 0: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n100-run0.json
  run 1: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n100-run1.json

A0-threshold / n150  [DELIVERED: 2/2]
  run 0: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n150-run0.json
  run 1: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n150-run1.json

A0-threshold / n160  [DELIVERED: 2/2]
  run 0: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n160-run0.json
  run 1: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n160-run1.json

A0-threshold / n163  [DELIVERED: 2/2]
  run 0: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n163-run0.json
  run 1: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n163-run1.json

A0-threshold / n164  [REJECTED: 2/2]
  run 0: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n164-run0.json
  run 1: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n164-run1.json

A0-threshold / n165  [REJECTED: 2/2]
  run 0: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n165-run0.json
  run 1: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n165-run1.json

A0-threshold / n170  [REJECTED: 2/2]
  run 0: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n170-run0.json
  run 1: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n170-run1.json

A0-threshold / n200  [REJECTED: 2/2]
  run 0: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n200-run0.json
  run 1: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/sdk-0.10.0/A0-threshold/n200-run1.json

C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2115,t=2950,t=2956
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2949,t=2953
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-reload-run1.json

C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2119,t=2954,t=2960
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2956,t=2962
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-navigate-run1.json

C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 2/2]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2124,t=2931,t=2936
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2134,t=2949,t=2959
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/inject503-then-close-run1.json

C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 2/2]
  run 0: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2960
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run0.json
  run 1: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2954
      evidence: ../evidence/sdk-0.10.0/C-unload-during-cooldown/control-no-failure-reload-run1.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5114ms); D-tail delivered=false; reload at t=1383ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5124ms); D-tail delivered=false; reload at t=1380ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9123ms); D-tail delivered=false; reload at t=5383ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9116ms); D-tail delivered=false; reload at t=5370ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json

D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 2/2]
  run 0: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8113ms); D-tail delivered=false; reload at t=6231ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6241ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json

D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 2/2]
  run 0: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1145ms); D-tail delivered=true (received t=1408ms); reload at t=1401ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
  run 1: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1115ms); D-tail delivered=true (received t=1396ms); reload at t=1390ms
      evidence: ../evidence/sdk-0.10.0/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json

D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
  run 0: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1161ms): span export delay max 14983 ms / min 886 ms; log export delay max 14983 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17144:1sp/1157B 18144:1sp/1157B 19144:1sp/1157B]
      evidence: ../evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
  run 1: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1131ms): span export delay max 14992 ms / min 922 ms; log export delay max 14992 ms (logs endpoint itself is fast); traces POSTs at server t=[1123:1sp/1160B 16128:15sp/6677B 17123:1sp/1157B 18123:1sp/1157B 19122:1sp/1157B]
      evidence: ../evidence/sdk-0.10.0/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2732] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2732B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 2/2]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6407B) sessionEvents(21371B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 2/2]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6416] events[508] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6416B) sessionEvents(508B) traces(60652B) sessionReplays/meta(1098B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 2/2]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1158] events[2133] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1158B) sessionEvents(2133B) sessionReplays/meta(1088B)
      evidence: ../evidence/sdk-0.10.0/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1777; requests issued around the reload (Playwright): traces(1165B)@1789 logs(1014B)@1790 sessionReplays/meta(1194B)@1790 sessionEvents(1043B)@1790 sessionReplays/meta(1090B)@1850; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1783 sessionReplays/meta@1785 logs@1785 sessionEvents@1787 sessionReplays/meta@1849; console: [MapleClientSDK] logs flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush failed; cooldown 60s@1790 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1790 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1790 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1790 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1790
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1807; requests issued around the reload (Playwright): traces(1165B)@1819 logs(1014B)@1819 sessionReplays/meta(1194B)@1819 sessionEvents(1043B)@1819 sessionReplays/meta(1090B)@1882; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: traces@1813 logs@1814 sessionReplays/meta@1815 sessionEvents@1816 sessionReplays/meta@1882; console: [MapleClientSDK] logs flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush failed; cooldown 60s@1820 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1820 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1820 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1819 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1819
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-on(default)-run1.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 2/2]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1783; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1792 sessionEvents(510B)@1793 traces(1165B)@1793 logs(1014B)@1793 sessionReplays/meta(1100B)@1825; Playwright failed: sessionReplays/meta[net::ERR_ABORTED] traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1789 sessionEvents@1790 traces@1790 logs@1791 sessionReplays/meta@1822; console: [MapleClientSDK] traces flush failed; cooldown 60s@1793 | [MapleClientSDK] logs flush failed; cooldown 60s@1793 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1793 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1793 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1793
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1772; requests issued around the reload (Playwright): sessionReplays/meta(1204B)@1782 sessionEvents(510B)@1783 traces(1165B)@1783 logs(1014B)@1784 sessionReplays/meta(1100B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1779 sessionEvents@1780 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1784 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1784 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1784 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
      evidence: ../evidence/sdk-0.10.0/K-reload-150ms-after-action/replay-off-run1.json

@@@@@@@@@@ END FILE: evidence/sdk-0.10.0/summary-shell-sdk0.10.0-2026-10-07T23-43-48-971Z.txt

@@@@@@@@@@ FILE: evidence/summary-shell-2026-10-07T23-39-09-001Z.txt (46826 bytes)
Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:39:09.001Z, finished 2026-10-07T23:58:20.324Z

B-probe / probe  [MEASURED: 2/2]
  run 0: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/B-keepalive-probe/probe-run0.json
  run 1: MEASURED — single: 1024:ok 49152:ok 65535:ok 65536:ok 65537:REJECTED 70000:REJECTED 131072:REJECTED | 2x40000 concurrent: ok/REJECTED | 4x20000 concurrent: ok/ok/ok/REJECTED | unload 2x40000 arrived 1/2, 4x20000 arrived 3/4 | B10 unload-while-inflight arrived B10-periodic-hold50000
      evidence: ../evidence/B-keepalive-probe/probe-run1.json

B-sdk-concurrent-signals / manual-flush  [REPRODUCED (one signal rejected): 3/3]
  run 0: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run1.json
  run 2: REPRODUCED (one signal rejected) — trigger=flush; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] logs flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/manual-flush-run2.json

B-sdk-concurrent-signals / reload  [REPRODUCED (one signal rejected): 3/3]
  run 0: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run0.json
  run 1: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] metrics flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run1.json
  run 2: REPRODUCED (one signal rejected) — trigger=reload; spans 40/40, logs 0/30, metrics POSTs 1; request bodies seen by Playwright [traces:38472 logs:32135 metrics:931]; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/B-sdk-concurrent-signals/reload-run2.json

A0-threshold / n100  [DELIVERED: 1/1]
  run 0: DELIVERED — 100 spans in one flush: arrived 100; traces POST body bytes (Playwright) 40252; console: none
      evidence: ../evidence/A0-threshold/n100-run0.json

A0-threshold / n150  [DELIVERED: 1/1]
  run 0: DELIVERED — 150 spans in one flush: arrived 150; traces POST body bytes (Playwright) 60052; console: none
      evidence: ../evidence/A0-threshold/n150-run0.json

A0-threshold / n160  [DELIVERED: 1/1]
  run 0: DELIVERED — 160 spans in one flush: arrived 160; traces POST body bytes (Playwright) 64012; console: none
      evidence: ../evidence/A0-threshold/n160-run0.json

A0-threshold / n163  [DELIVERED: 1/1]
  run 0: DELIVERED — 163 spans in one flush: arrived 163; traces POST body bytes (Playwright) 65200; console: none
      evidence: ../evidence/A0-threshold/n163-run0.json

A0-threshold / n164  [REJECTED: 1/1]
  run 0: REJECTED — 164 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65596; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n164-run0.json

A0-threshold / n165  [REJECTED: 1/1]
  run 0: REJECTED — 165 spans in one flush: arrived 0; traces POST body bytes (Playwright) 65992; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n165-run0.json

A0-threshold / n170  [REJECTED: 1/1]
  run 0: REJECTED — 170 spans in one flush: arrived 0; traces POST body bytes (Playwright) 67972; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n170-run0.json

A0-threshold / n200  [REJECTED: 1/1]
  run 0: REJECTED — 200 spans in one flush: arrived 0; traces POST body bytes (Playwright) 79852; console: [MapleClientSDK] traces flush failed; cooldown 60s: TypeError: Failed to fetch
      evidence: ../evidence/A0-threshold/n200-run0.json

C-unload-during-cooldown / inject503-then-reload  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2937,t=2942
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2947,t=2956
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=reload: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2122,t=2952,t=2958
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-reload-run2.json

C-unload-during-cooldown / inject503-then-navigate  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2937,t=2941
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2108,t=2937,t=2943
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=navigate: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2147,t=2960,t=2965
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-navigate-run2.json

C-unload-during-cooldown / inject503-then-close  [REPRODUCED (both spans lost): 3/3]
  run 0: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2116,t=2938,t=2942
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run0.json
  run 1: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2120,t=2930,t=2935
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run1.json
  run 2: REPRODUCED (both spans lost) — inject503=true unload=close: C-first delivered=false, C-before-unload delivered=false; console failed=1 skipped=t=2111,t=2939,t=2945
      evidence: ../evidence/C-unload-during-cooldown/inject503-then-close-run2.json

C-unload-during-cooldown / control-no-failure-reload  [CONTROL OK (both delivered): 3/3]
  run 0: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2972
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run0.json
  run 1: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2949
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run1.json
  run 2: CONTROL OK (both delivered) — inject503=false unload=reload: C-first delivered=true, C-before-unload delivered=true; console failed=1 skipped=t=2970
      evidence: ../evidence/C-unload-during-cooldown/control-no-failure-reload-run2.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval1000  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5125ms); D-tail delivered=false; reload at t=1401ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5107ms); D-tail delivered=false; reload at t=1382ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=5108ms); D-tail delivered=false; reload at t=1375ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval1000-run2.json

D-pagehide-behind-inflight-flush / first-POST-4000ms-interval5000  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9124ms); D-tail delivered=false; reload at t=5383ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9115ms); D-tail delivered=false; reload at t=5367ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 4000 ms, autoFlushInterval 5000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=9114ms); D-tail delivered=false; reload at t=5373ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-4000ms-interval5000-run2.json

D-pagehide-behind-inflight-flush / first-POST-7000ms-interval1000-tail-5s-before-reload  [REPRODUCED (tail span lost): 3/3]
  run 0: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8107ms); D-tail delivered=false; reload at t=6226ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run0.json
  run 1: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8114ms); D-tail delivered=false; reload at t=6240ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run1.json
  run 2: REPRODUCED (tail span lost) — first traces POST held 7000 ms, autoFlushInterval 1000, D-tail ended 5000 ms before reload: D-early delivered=true (server responded t=8150ms); D-tail delivered=false; reload at t=6268ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/first-POST-7000ms-interval1000-tail-5s-before-reload-run2.json

D-pagehide-behind-inflight-flush / control-no-latency-interval1000  [CONTROL OK: 3/3]
  run 0: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1405ms); reload at t=1393ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run0.json
  run 1: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1134ms); D-tail delivered=true (received t=1412ms); reload at t=1406ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run1.json
  run 2: CONTROL OK — first traces POST held 0 ms, autoFlushInterval 1000, D-tail ended 150 ms before reload: D-early delivered=true (server responded t=1123ms); D-tail delivered=true (received t=1401ms); reload at t=1394ms
      evidence: ../evidence/D-pagehide-behind-inflight-flush/control-no-latency-interval1000-run2.json

D-hang-head-of-line-blocking / first-POST-held-15s  [REPRODUCED (exports stalled behind one slow request): 2/2]
  run 0: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1153ms): span export delay max 14990 ms / min 909 ms; log export delay max 14991 ms (logs endpoint itself is fast); traces POSTs at server t=[1145:1sp/1160B 16149:15sp/6677B 17143:1sp/1157B 18143:1sp/1157B 19144:1sp/1157B]
      evidence: ../evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run0.json
  run 1: REPRODUCED (exports stalled behind one slow request) — first traces POST held 15000 ms (from t=1137ms): span export delay max 14977 ms / min 905 ms; log export delay max 14977 ms (logs endpoint itself is fast); traces POSTs at server t=[1116:1sp/1160B 16120:15sp/6677B 17115:1sp/1157B 18115:1sp/1157B 19115:1sp/1157B]
      evidence: ../evidence/D-hang-head-of-line-blocking/first-POST-held-15s-run1.json

Dr-random-reload-latency / latency150-interval1000  [no loss: 9/10; LOSS (1 spans): 1/10]
  run 0: no loss — latency 150 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run0.json
  run 1: no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run1.json
  run 2: no loss — latency 150 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run2.json
  run 3: no loss — latency 150 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run3.json
  run 4: no loss — latency 150 ms, interval 1000: created 38, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run4.json
  run 5: no loss — latency 150 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run5.json
  run 6: no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run6.json
  run 7: LOSS (1 spans) — latency 150 ms, interval 1000: created 40, lost 1 (created 100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run7.json
  run 8: no loss — latency 150 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run8.json
  run 9: no loss — latency 150 ms, interval 1000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency150-interval1000-run9.json

Dr-random-reload-latency / latency600-interval1000  [no loss: 7/10; LOSS (5 spans): 1/10; LOSS (2 spans): 1/10; LOSS (4 spans): 1/10]
  run 0: no loss — latency 600 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run0.json
  run 1: no loss — latency 600 ms, interval 1000: created 46, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run1.json
  run 2: LOSS (5 spans) — latency 600 ms, interval 1000: created 54, lost 5 (created 513,411,305,202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run2.json
  run 3: LOSS (2 spans) — latency 600 ms, interval 1000: created 22, lost 2 (created 202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run3.json
  run 4: no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run4.json
  run 5: no loss — latency 600 ms, interval 1000: created 57, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run5.json
  run 6: no loss — latency 600 ms, interval 1000: created 39, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run6.json
  run 7: LOSS (4 spans) — latency 600 ms, interval 1000: created 34, lost 4 (created 406,304,202,100 ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run7.json
  run 8: no loss — latency 600 ms, interval 1000: created 28, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run8.json
  run 9: no loss — latency 600 ms, interval 1000: created 47, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval1000-run9.json

Dr-random-reload-latency / latency600-interval5000  [no loss: 10/10]
  run 0: no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run0.json
  run 1: no loss — latency 600 ms, interval 5000: created 41, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run1.json
  run 2: no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run2.json
  run 3: no loss — latency 600 ms, interval 5000: created 22, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run3.json
  run 4: no loss — latency 600 ms, interval 5000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: true
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run4.json
  run 5: no loss — latency 600 ms, interval 5000: created 42, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run5.json
  run 6: no loss — latency 600 ms, interval 5000: created 35, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run6.json
  run 7: no loss — latency 600 ms, interval 5000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run7.json
  run 8: no loss — latency 600 ms, interval 5000: created 21, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run8.json
  run 9: no loss — latency 600 ms, interval 5000: created 25, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/latency600-interval5000-run9.json

Dr-random-reload-latency / control-latency0-interval1000  [no loss: 10/10]
  run 0: no loss — latency 0 ms, interval 1000: created 49, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run0.json
  run 1: no loss — latency 0 ms, interval 1000: created 58, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run1.json
  run 2: no loss — latency 0 ms, interval 1000: created 43, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run2.json
  run 3: no loss — latency 0 ms, interval 1000: created 26, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run3.json
  run 4: no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run4.json
  run 5: no loss — latency 0 ms, interval 1000: created 31, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run5.json
  run 6: no loss — latency 0 ms, interval 1000: created 54, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run6.json
  run 7: no loss — latency 0 ms, interval 1000: created 33, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run7.json
  run 8: no loss — latency 0 ms, interval 1000: created 44, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run8.json
  run 9: no loss — latency 0 ms, interval 1000: created 23, lost 0 (created  ms before reload); trace flush in flight at reload: false
      evidence: ../evidence/Dr-random-reload-latency/control-latency0-interval1000-run9.json

E-combined-unload-budget / E1-spans45-plus-60kB-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 45/45; old-page unload arrivals (bytes): traces[18707] meta-ended[2702] events[] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: traces(18707B) sessionReplays/meta(2702B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E1-spans45-plus-60kB-DOM-mutation-run2.json

E-combined-unload-budget / E2-spans150(~60kB)-plus-30-console-lines  [LOST: meta(ended), final sessionEvents: 3/3]
  run 0: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run0.json
  run 1: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run1.json
  run 2: LOST: meta(ended), final sessionEvents — spans 150/150; old-page unload arrivals (bytes): traces[60652] meta-ended[] events[] blob[]; replay chunkSeqs at server [0,1]; Playwright during unload: traces(60652B) sessionReplays/meta(6377B) sessionEvents(21041B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E2-spans150(~60kB)-plus-30-console-lines-run2.json

E-combined-unload-budget / E3-replay-disabled-spans150(~60kB)  [LOST: traces: 3/3]
  run 0: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run0.json
  run 1: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run1.json
  run 2: LOST: traces — spans 0/150; old-page unload arrivals (bytes): traces[] meta-ended[6386] events[497] blob[]; replay chunkSeqs at server []; Playwright during unload: sessionReplays/meta(6386B) sessionEvents(497B) traces(60652B) sessionReplays/meta(1068B)
      evidence: ../evidence/E-combined-unload-budget/E3-replay-disabled-spans150(~60kB)-run2.json

E-combined-unload-budget / E4-no-spans-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E4-no-spans-small-DOM-mutation-run2.json

E-combined-unload-budget / E5-Maple.layer-small-DOM-mutation  [LOST: replay chunk(s) 1 (seq consumed, never posted): 3/3]
  run 0: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run0.json
  run 1: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run1.json
  run 2: LOST: replay chunk(s) 1 (seq consumed, never posted) — spans 0/0; old-page unload arrivals (bytes): traces[] meta-ended[1128] events[2100] blob[]; replay chunkSeqs at server [0,2]; Playwright during unload: sessionReplays/meta(1128B) sessionEvents(2100B) sessionReplays/meta(1058B)
      evidence: ../evidence/E-combined-unload-budget/E5-Maple.layer-small-DOM-mutation-run2.json

F-replay-posts-aborted-by-navigation / flush-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run0.json
  run 1: REPRODUCED — flush/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-periodic-events-POST-in-flight-run1.json

F-replay-posts-aborted-by-navigation / flush-active-meta-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run0.json
  run 1: REPRODUCED — flush/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/flush-active-meta-POST-in-flight-run1.json

Fr-random-reloads / flush-8-random-reloads-latency300  [REPRODUCED: 2/2]
  run 0: REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"traces[net::ERR_ABORTED]":1,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: ../evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run0.json
  run 1: REPRODUCED — flush, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST; [maple] session replay blob PUT], SDK flush failed 1, skipped 1; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":5,"sessionReplays/blob[net::ERR_ABORTED]":1,"traces[net::ERR_ABORTED]":1}
      evidence: ../evidence/Fr-random-reloads/flush-8-random-reloads-latency300-run1.json

F-replay-posts-aborted-by-navigation / layer-periodic-events-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run0.json
  run 1: REPRODUCED — layer/events: console: [maple] session replay metadata POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-periodic-events-POST-in-flight-run1.json

F-replay-posts-aborted-by-navigation / layer-active-meta-POST-in-flight  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run0.json
  run 1: REPRODUCED — layer/meta: console: [maple] session replay events POST failed (dropping; will retry on next chunk): TypeError: Failed to fetch
      evidence: ../evidence/F-replay-posts-aborted-by-navigation/layer-active-meta-POST-in-flight-run1.json

Fr-random-reloads / layer-8-random-reloads-latency300  [REPRODUCED: 2/2]
  run 0: REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 15 [[maple] session replay events POST; [maple] session replay metadata POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionEvents[net::ERR_ABORTED]":7,"sessionReplays/meta[net::ERR_ABORTED]":8}
      evidence: ../evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run0.json
  run 1: REPRODUCED — layer, latency 300 ms, 8 reloads: replay-post console errors 14 [[maple] session replay metadata POST; [maple] session replay events POST], SDK flush failed 0, skipped 0; Playwright requestfailed by path: {"sessionReplays/meta[net::ERR_ABORTED]":8,"sessionEvents[net::ERR_ABORTED]":6}
      evidence: ../evidence/Fr-random-reloads/layer-8-random-reloads-latency300-run1.json

G-playwright-keepalive-artifact / steady-state-no-navigation  [no requestfailed events: 2/2]
  run 0: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run0.json
  run 1: no requestfailed events — reload=false: traces: pw req 8/fin 8/fail 0 vs server200 8; sessionReplays/meta: pw req 1/fin 1/fail 0 vs server200 1; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/steady-state-no-navigation-run1.json

G-playwright-keepalive-artifact / with-final-reload  [ARTIFACT (Playwright reports failures for requests the server answered 200): 2/2]
  run 0: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/with-final-reload-run0.json
  run 1: ARTIFACT (Playwright reports failures for requests the server answered 200) — reload=true: traces: pw req 9/fin 8/fail 1[net::ERR_ABORTED] vs server200 9; sessionReplays/meta: pw req 3/fin 2/fail 1[net::ERR_ABORTED] vs server200 3; sessionEvents: pw req 1/fin 1/fail 0 vs server200 1; sessionReplays/blob: pw req 1/fin 1/fail 0 vs server200 1; logs: pw req 0/fin 0/fail 0 vs server200 0; metrics: pw req 0/fin 0/fail 0 vs server200 0
      evidence: ../evidence/G-playwright-keepalive-artifact/with-final-reload-run1.json

H-maple-layer-unload / layer  [tail LOST: 3/3]
  run 0: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run0.json
  run 1: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run1.json
  run 2: tail LOST — layer: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=false
      evidence: ../evidence/H-maple-layer-unload/layer-run2.json

H-maple-layer-unload / layer-flusher  [tail delivered: 3/3]
  run 0: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run0.json
  run 1: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run1.json
  run 2: tail delivered — layer-flusher: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/layer-flusher-run2.json

H-maple-layer-unload / flush  [tail delivered: 3/3]
  run 0: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run0.json
  run 1: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run1.json
  run 2: tail delivered — flush: H-early (created 7.5 s before reload) delivered=true; H-tail (created 1 s before reload) delivered=true
      evidence: ../evidence/H-maple-layer-unload/flush-run2.json

I-cross-origin / same-origin  [tail delivered: 2/2]
  run 0: tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/same-origin-run0.json
  run 1: tail delivered — same-origin: I-tail delivered=true; preflights 0 (access-control-request-headers: n/a); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/same-origin-run1.json

I-cross-origin / cross-origin  [tail delivered: 2/2]
  run 0: tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/cross-origin-run0.json
  run 1: tail delivered — cross-origin: I-tail delivered=true; preflights 5 (access-control-request-headers: authorization,content-type,x-maple-sdk); traces POST user-agent header: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, l, x-maple-sdk: maple-effect-sdk-client/0.8.2
      evidence: ../evidence/I-cross-origin/cross-origin-run1.json

K-reload-150ms-after-action / replay-on(default)  [span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed]: 3/3]
  run 0: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1795; requests issued around the reload (Playwright): traces(1165B)@1807 logs(1014B)@1807 sessionReplays/meta(1164B)@1807 sessionEvents(1021B)@1807 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1801 logs@1802 sessionReplays/meta@1802 sessionEvents@1804 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush failed; cooldown 60s@1808 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1808 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1808 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1808 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1808
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run0.json
  run 1: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1810; requests issued around the reload (Playwright): traces(1165B)@1824 logs(1014B)@1824 sessionReplays/meta(1164B)@1824 sessionEvents(1021B)@1824 sessionReplays/meta(1060B)@1891; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1817 sessionReplays/meta@1818 logs@1818 sessionEvents@1820 sessionReplays/meta@1890; console: [MapleClientSDK] logs flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush failed; cooldown 60s@1825 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1825 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1825 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1825 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1825
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run1.json
  run 2: span delivered, log delivered; console flush-failed: [logs flush failed, traces flush failed] — reload at t=1803; requests issued around the reload (Playwright): traces(1165B)@1815 logs(1014B)@1816 sessionReplays/meta(1164B)@1816 sessionEvents(1021B)@1816 sessionReplays/meta(1060B)@1871; Playwright failed: logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] traces[net::ERR_ABORTED]; server 200 in same window: traces@1809 logs@1810 sessionReplays/meta@1813 sessionEvents@1814 sessionReplays/meta@1870; console: [MapleClientSDK] logs flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush failed; cooldown 60s@1816 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1816 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1816 | [maple] session replay metadata POST failed (dropping; will retry on next chunk)@1816 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1816
      evidence: ../evidence/K-reload-150ms-after-action/replay-on(default)-run2.json

K-reload-150ms-after-action / replay-off  [span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed]: 3/3]
  run 0: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1770; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1781 sessionEvents(499B)@1781 traces(1165B)@1782 logs(1014B)@1782 sessionReplays/meta(1070B)@1814; Playwright failed: traces[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1777 traces@1778 sessionEvents@1779 logs@1781 sessionReplays/meta@1811; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1787 | [MapleClientSDK] traces flush skipped (cooldown 60000ms remaining)@1787 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1788 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1782
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run0.json
  run 1: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1771; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1783 sessionEvents(499B)@1783 traces(1165B)@1784 logs(1014B)@1784 sessionReplays/meta(1070B)@1817; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1778 sessionEvents@1779 traces@1780 logs@1781 sessionReplays/meta@1814; console: [MapleClientSDK] traces flush failed; cooldown 60s@1784 | [MapleClientSDK] logs flush failed; cooldown 60s@1785 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1785 | [MapleClientSDK] logs flush skipped (cooldown 59999ms remaining)@1785 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1784
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run1.json
  run 2: span delivered, log delivered; console flush-failed: [traces flush failed, logs flush failed] — reload at t=1769; requests issued around the reload (Playwright): sessionReplays/meta(1174B)@1780 sessionEvents(499B)@1780 traces(1165B)@1781 logs(1014B)@1781 sessionReplays/meta(1070B)@1813; Playwright failed: traces[net::ERR_ABORTED] sessionEvents[net::ERR_ABORTED] logs[net::ERR_ABORTED] sessionReplays/meta[net::ERR_ABORTED]; server 200 in same window: sessionReplays/meta@1776 traces@1777 sessionEvents@1778 logs@1780 sessionReplays/meta@1810; console: [MapleClientSDK] traces flush failed; cooldown 60s@1782 | [MapleClientSDK] logs flush failed; cooldown 60s@1782 | [MapleClientSDK] traces flush skipped (cooldown 59999ms remaining)@1782 | [MapleClientSDK] logs flush skipped (cooldown 60000ms remaining)@1782 | [maple] session replay events POST failed (dropping; will retry on next chunk)@1781
      evidence: ../evidence/K-reload-150ms-after-action/replay-off-run2.json

L-multi-arg-log-body / logError-4-args  [body is an OTLP arrayValue (structured), not a JSON string: 1/1]
  run 0: body is an OTLP arrayValue (structured), not a JSON string — OTLP logRecord.body as POSTed by the SDK: {"arrayValue":{"values":[{"stringValue":"L Failed to connect"},{"stringValue":"Error: ClientError: boom"},{"stringValue":"Unknown Error"},{"stringValue":"{\"code\":42,\"nested\":{\"a\":1}}"}]}}
      evidence: ../evidence/L-multi-arg-log-body/logError-4-args-run0.json

@@@@@@@@@@ END FILE: evidence/summary-shell-2026-10-07T23-39-09-001Z.txt

@@@@@@@@@@ FILE: evidence/summary-shell-2026-10-07T23-39-09-004Z.txt (3061 bytes)
Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:39:09.004Z, finished 2026-10-07T23:41:28.743Z

A-oversized-batch / steady30spansPerSec-10s-interval5000(default)  [NOT REPRODUCED: 2/2]
  run 1: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run1.json
  run 0: NOT REPRODUCED — created 338 spans (38 small 'tick' spans after the burst); delivered 338 (ticks 38/38); traces POSTs reaching server: 18; logs delivered 8; console 'flush failed' at t=ms
      evidence: ../evidence/A-oversized-batch/steady30spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / steady40spansPerSec-10s-interval5000(default)  [PARTIAL LOSS: 2/2]
  run 1: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: ../evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run1.json
  run 0: PARTIAL LOSS — created 438 spans (38 small 'tick' spans after the burst); delivered 152 (ticks 0/38); traces POSTs reaching server: 1; logs delivered 8; console 'flush failed' at t=10506,75510ms
      evidence: ../evidence/A-oversized-batch/steady40spansPerSec-10s-interval5000(default)-run0.json

A-oversized-batch / burst200x500B-interval1000  [REPRODUCED (no span ever delivered): 2/2]
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2565,63567,124567ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval1000-run1.json
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=2548,63547,124547ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval1000-run0.json

A-oversized-batch / burst200x500B-interval5000(default)  [REPRODUCED (no span ever delivered): 2/2]
  run 0: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5491,70488,135493ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval5000(default)-run0.json
  run 1: REPRODUCED (no span ever delivered) — created 265 spans (65 small 'tick' spans after the burst); delivered 0 (ticks 0/65); traces POSTs reaching server: 0; logs delivered 13; console 'flush failed' at t=5483,70483,135481ms
      evidence: ../evidence/A-oversized-batch/burst200x500B-interval5000(default)-run1.json

@@@@@@@@@@ END FILE: evidence/summary-shell-2026-10-07T23-39-09-004Z.txt

@@@@@@@@@@ FILE: evidence/summary-shell-2026-10-07T23-58-29-444Z.txt (1271 bytes)
Scenario summary (chromium-headless-shell (Playwright default), Chromium 143.0.7499.4, Playwright 1.57.0, Bun 1.4.2)
started 2026-10-07T23:58:29.444Z, finished 2026-10-07T23:59:04.742Z

W-workaround-keepalive-downgrade-shim / without-shim  [LOSS (burst 0/200, ticks 0/5): 2/2]
  run 0: LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/without-shim-run0.json
  run 1: LOSS (burst 0/200, ticks 0/5) — shim=false: burst spans delivered 0/200, later spans 0/5; console flush-failed 1, skipped 6
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/without-shim-run1.json

W-workaround-keepalive-downgrade-shim / with-keepalive-downgrade-shim  [all delivered: 2/2]
  run 0: all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run0.json
  run 1: all delivered — shim=true: burst spans delivered 200/200, later spans 5/5; console flush-failed 0, skipped 0
      evidence: ../evidence/W-workaround-keepalive-downgrade-shim/with-keepalive-downgrade-shim-run1.json

@@@@@@@@@@ END FILE: evidence/summary-shell-2026-10-07T23-58-29-444Z.txt

@@@@@@@@@@ FILE: evidence/webkit-firefox-launch.txt (1867 bytes)
webkit launch failed:
Error: launch: 
╔══════════════════════════════════════════════════════╗
║ Host system is missing dependencies to run browsers. ║
║ Please install them with the following command:      ║
║                                                      ║
║     sudo npx playwright install-deps                 ║
║                                                      ║
║ Alternatively, use apt:                              ║
║     sudo apt-get install libicu74\                   ║
║         libatomic1\                                  ║
║         libvpx9\                                     ║
║         libjpeg-turbo8\                              ║
firefox launch failed:
Error: launch: Executable doesn't exist at ~/.cache/ms-playwright/firefox-1497/firefox/firefox
╔═════════════════════════════════════════════════════════════════════════╗
║ Looks like Playwright Test or Playwright was just installed or updated. ║
║ Please run the following command to download new browsers:              ║
║                                                                         ║
║     npx playwright install                                              ║
║                                                                         ║
║ <3 Playwright Team                                                      ║
╚═════════════════════════════════════════════════════════════════════════╝
@@@@@@@@@@ END FILE: evidence/webkit-firefox-launch.txt