[kixelated]

Progress on quest/m1/audio-jitter-target/capped-budgets.md. Five of the seven scheduled nightlies since 2026-10-25 have run. budgets.json is unchanged: the Plan says to re-measure only from seven nights, and the first four of these measured a relay bug.

Findings

• 2026-10-25 to 10-28 measured a relay history-fill bug, not the player. 2 to 4.5 s into every browser row except fixed-250 and Opus near-zero, the subscriber received 15 to 25 s of the publisher's cached history, newest first. On the wide rows the estimator treated it as a late interval, held a 2020 ms target for about 20 s, then discarded 1.8 to 1.9 s of audio a minute falling back. That happened on all four nights, so it is the 10-28 AAC/Opus wide breach. The 10-28 Opus bursty 540 ms target lines up with the same arrival at 4.5 s. No wide or bursty row logged more than one long task (at most 58 ms), so these are not runner stalls.
• moq-dev/moq#7168 fixed it (a fresh subscriber no longer fills history below its start). It merged 2026-10-29T00:52Z, before the 2026-10-29 nightly. On that night the oldest frame any row received was about 2.6 s behind live, and every browser row passed: AAC wide discarded 205.6 ms/min at a 323.2 ms target p95, and Opus wide 243.4 at 460.
• Quest goal, so far: across all five nights, AAC bursty discarded 0 to 238.1 ms/min against Opus bursty's 0 to 370.9. Opus near-zero had no underrun episodes. Its target p95 was 60 ms on three nights and 80 on two, because the 120 ms prior only decays to 60 at 21.4 s, after the 5 s warmup. The Plan says to budget that startup hold as measured.

Changes

• quest/m1/audio-jitter-target/capped-budgets.md: records the history-fill finding, the restarted count (1 of 7, from 2026-10-29), and the 2026-10-29 goal readings, and points here for the per-night data. I did not touch budgets.json.

Decisions

1. ✅ Restart the count at 2026-10-29 (maintainer, 2026-10-30), the first nightly after moq-dev/moq#7168. The 2026-10-25 to 10-28 nightlies no longer count. One of seven has run; the seventh is due about 2026-11-04. The quest's Goal and Plan now say so.

Follow-ups

• Possible quest: should the estimator count a frame that arrives already behind the playhead (stale history) as an arrival interval? Here one burst of old frames held a 2 s target for 20 s. With moq-dev/moq#7168 the relay no longer sends that history, but a publisher or relay that does would hit the same thing.

Checks

• quest check --base origin/main: passed on f4bf5e8f7a.
• just check: not rerun on f4bf5e8f7a; that commit only edits quest prose.
• quest gate just check: passed (markdown, gh, quest, shell, toml, nix, justfile).

Review

• Head f4bf5e8f7a8a7fa917ddb2d0e84a2af4798dc3d4: review skipped as trivial. The commit only records the maintainer's restart decision in the quest prose.
• Earlier round: Codex, gpt-6.1-sol, medium effort (quest review --base origin/main, round 1), on 8add6f8e107849e2fa4ef1f05ed7ce34830fd7ed. No findings.

Per-night browser metrics, 2026-10-25 to 10-29 (the artifacts expire after 7 days)

Runs: 10-25 40192013303, 10-26 40332938121, 10-27 40506938673, 10-28 40698329026, 10-29 40885347517. No row voided.

row	night	underruns/min	gap ms/min	skips/min	discarded ms/min	stalled	silence	target p95
aac-44100-bursty	10-25	2.2	5.2	0	0	0	0	323.2
aac-44100-bursty	10-26	0	0	2.2	238.1	0	0	683.2
aac-44100-bursty	10-27	0	0	2.2	234.6	0.005	0.003	343.2
aac-44100-bursty	10-28	0	0	0	0	0	0.003	343.2
aac-44100-bursty	10-29	0	0	1.1	162.9	0	0	323.2
aac-44100-fixed-250	10-25	0	0	0	0	0	0	250
aac-44100-fixed-250	10-26	0	0	0	0	0	0	250
aac-44100-fixed-250	10-27	0	0	0	0	0	0	250
aac-44100-fixed-250	10-28	0	0	0	0	0	0	250
aac-44100-fixed-250	10-29	0	0	0	0	0	0	250
aac-44100-mild	10-25	0	0	0	0	0	0	209.2
aac-44100-mild	10-26	0	0	0	0	0	0	209.2
aac-44100-mild	10-27	0	0	0	0	0	0	209.2
aac-44100-mild	10-28	0	0	0	0	0	0	209.2
aac-44100-mild	10-29	0	0	0	0	0	0	209.2
aac-44100-near-zero	10-25	0	0	0	0	0	0	209.2
aac-44100-near-zero	10-26	0	0	0	0	0	0	223.2
aac-44100-near-zero	10-27	0	0	0	0	0	0	223.2
aac-44100-near-zero	10-28	0	0	0	0	0	0	209.2
aac-44100-near-zero	10-29	0	0	0	0	0	0	209.2
aac-44100-step	10-25	0	0	0	0	0	0.006	223.2
aac-44100-step	10-26	1.1	0.3	0	0	0.005	0.003	223.2
aac-44100-step	10-27	1.1	2.9	0	0	0	0.006	223.2
aac-44100-step	10-28	3.3	9.5	0	0	0	0.003	223.2
aac-44100-step	10-29	1.1	2.9	0	0	0.005	0.006	223.2
aac-44100-wide	10-25	0	0	2.2	1820.9	0.005	0.003	2023.2
aac-44100-wide	10-26	0	0	2.2	1803.9	0.005	0.004	2023.2
aac-44100-wide	10-27	0	0	2.2	1859.1	0.005	0.004	2023.2
aac-44100-wide	10-28	1.1	2	2.2	1790.9	0	0.002	2023.2
aac-44100-wide	10-29	1.1	1.7	4.4	205.6	0.018	0.012	323.2
opus-48000-bursty	10-25	0	0	0	0	0	0	180
opus-48000-bursty	10-26	0	0	0	0	0	0	200
opus-48000-bursty	10-27	0	0	0	0	0	0	220
opus-48000-bursty	10-28	0	0	1.1	370.9	0	0	540
opus-48000-bursty	10-29	0	0	0	0	0	0	180
opus-48000-fixed-250	10-25	0	0	0	0	0	0	250
opus-48000-fixed-250	10-26	0	0	0	0	0	0	250
opus-48000-fixed-250	10-27	0	0	0	0	0	0	250
opus-48000-fixed-250	10-28	0	0	0	0	0	0	250
opus-48000-fixed-250	10-29	0	0	0	0	0	0	250
opus-48000-mild	10-25	0	0	1.1	56	0	0	140
opus-48000-mild	10-26	2.2	3	0	0	0	0.001	120
opus-48000-mild	10-27	0	0	1.1	124.4	0	0	180
opus-48000-mild	10-28	1.1	1.5	1.1	124	0.005	0.002	160
opus-48000-mild	10-29	0	0	1.1	60.5	0	0.001	120
opus-48000-near-zero	10-25	0	0	0	0	0	0	60
opus-48000-near-zero	10-26	0	0	0	0	0	0	80
opus-48000-near-zero	10-27	0	0	0	0	0	0	60
opus-48000-near-zero	10-28	0	0	0	0	0	0	80
opus-48000-near-zero	10-29	0	0	0	0	0	0	60
opus-48000-step	10-25	0	0	1.1	91.5	0.005	0.002	160
opus-48000-step	10-26	0	0	1.1	90.3	0.009	0.002	160
opus-48000-step	10-27	0	0	2.2	171.8	0.005	0.002	180
opus-48000-step	10-28	0	0	2.2	98.3	0.005	0.002	120
opus-48000-step	10-29	0	0	0	0	0.005	0.002	100
opus-48000-wide	10-25	0	0	2.2	1972.4	0.005	0.001	2020
opus-48000-wide	10-26	0	0	4.4	1923.4	0.005	0.001	2020
opus-48000-wide	10-27	0	0	1.1	1902.7	0	0	2020
opus-48000-wide	10-28	0	0	3.3	1913.9	0	0	2020
opus-48000-wide	10-29	1.1	1.5	2.2	243.4	0	0	460

🤖 Generated with Claude Code

(Written by Claude Opus 5.5)

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 12 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: c4eca9fe-bfcb-48ea-8f08-a45f395af457

Commits

Reviewing files that changed from the base of the PR and between 1c733a0 and f4bf5e8.

Files selected for processing: 1

• quest/m1/audio-jitter-target/capped-budgets.md

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-30T20:14:39.306962Z	f4bf5e8	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: f4bf5e8

No actionable bugs found in this documentation-only change. Restarting the seven-night measurement window after the history-fill fix is a sound direction: it avoids calibrating budgets against the known over-fetch, preserves the earlier observations, and leaves the existing ceilings unchanged until enough post-fix data exists. No additional implementation complexity is introduced.

Verification: inspected the complete diff, surrounding quest plan, current budget rules/ceilings, and the referenced #7168 fix. No prior published review findings to carry forward. I did not run tests or independently reproduce the nightly trace analysis; the reported measurements are not independently validated here.