[kixelated]

Summary

Completes and deletes quest/m0/ts-duration-fidelity.md. The cause is already fixed on main by #5292, so this PR only records the evidence, removes the quest's references, and corrects an outdated comment.

Root cause

• Slow ingest. The stdin importer read 64 bytes per blocking-pool round trip. On the CI runner's debug build that could not keep up with the harness's 10 Mb/s PCR-paced source. In the failing main run (38058995817, c0c0bb33), the publisher took about 24 s to deliver the 20.06 s clip. The subscriber's release clock measured the source at -222,204 ppm.
• Late frames dropped. The fixed-delay export (feat(moq-mux)!: fixed-delay jitter buffer and per-PID T-STD admission for TS export #4645, merged 2026-10-08) drops any frame that misses its deadline. A source running about 22% slow uses up the delay within a few seconds, and after that every frame arrives late. That run dropped 1,273 frames and captured 3.2 s. fix(ts): schedule padded output at exact decoder times #5306's run 38118020915 shows the same signature: 2.5 s captured, -218,579 ppm. It ran before fix(cli): avoid tiny blocking stdin reads #5292 landed.
• Why it started when it did. The 64-byte reads date back to April. Before feat(moq-mux)!: fixed-delay jitter buffer and per-PID T-STD admission for TS export #4645, a slow ingest produced a stretched capture that still passed the band. That explains why the last green main run was 11b28de0 (2026-10-05) and the failures began once feat(moq-mux)!: fixed-delay jitter buffer and per-PID T-STD admission for TS export #4645 landed.

Evidence the fix holds

• Since fix(cli): avoid tiny blocking stdin reads #5292 merged (2026-10-11 16:07 UTC), 59 Interop runs reached TS compliance. Together they ran 353 duration-fidelity checks across the six TS arms, with 0 failures.
• The main runs on 2026-10-12 and 2026-10-13 pass duration-fidelity. Their remaining red comes from other quests, such as unroutable.
• fix(cli): avoid tiny blocking stdin reads #5292 added the regression stdin_reads_whole_available_packets, which fails without the fix and runs in just check. It also added a nightly import-throughput benchmark.

Remaining lead (accepted)

Since #5287 (2026-10-12), the ratio sits at about 0.926 instead of 1.021. A local capture starts at video PTS 3.48 against the source's 1.4, so exactly the first two GOPs are missing. Import holds the catalog for its two-second mux-rate window, and the exporter then joins at the live edge, so groups published during the hold are never read. This is within the band and was a stated cost of #5287. It does mean a subscriber waiting before the publisher no longer gets the whole stream. This PR updates the duration-fidelity docstring, which still claimed at most a GOP of lead. The maintainer accepted this gap as the cost of the catalog hold, so no follow-up is planned.

Conflicts

#5306 adds evidence to the deleted quest file. On rebase, that hunk should simply be dropped.

Also restores the trailing newline this PR dropped from quest/m1/ts-same-instance-resume.md.

Impact

• Public API: none.
• Wire: none.
• Code: a comment in test/ts/compliance.py, plus quest files.

Tests

• just check passes.
• quest check passes.
• just test ts passes locally twice on main: captured 18.58 s of 20.06 s.

🤖 Generated with Claude Code

(Written by Claude Opus 5.5)

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[kixelated]

Outcome: the quest's goal was met on main by #5292. No new code is needed beyond a docstring correction.

Open decision for the maintainer: should a TS exporter that is waiting before the publisher appears recover the roughly 2 s of groups published during #5287's catalog hold?

• Recommended: accept it as the hold's cost (this PR).
• Alternative: a follow-up quest to have the exporter read from the first group when it acquires a fresh broadcast.

(Written by Claude Opus 5.5)

[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 2026-10-13T19:01:45.278791Z	b24c3ad	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.

[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 8 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

• Configuration used: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 5592a9b3-764e-4e5d-aca7-1a686bbe401a

📥 Commits

Reviewing files that changed from the base of the PR and between 3b71e0e and b24c3ad.

📒 Files selected for processing: 5

• quest/m0/README.md
• quest/m0/ts-duration-fidelity.md
• quest/m1/ts-reliability/validation.md
• quest/m1/ts-same-instance-resume.md
• test/ts/compliance.py

• 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.}

[kixelated]

Merge summary:

• Deletes quest/m0/ts-duration-fidelity.md and removes its references. fix(cli): avoid tiny blocking stdin reads #5292 already fixed the cause: 353 duration-fidelity checks across 59 Interop runs since that merge, with 0 failures.
• Updates the check_duration_fidelity docstring to describe the real lead: the catalog hold's two-second window plus up to one GOP.
• Restores the trailing newline this PR had dropped from quest/m1/ts-same-instance-resume.md.

Decision: the maintainer accepted the roughly 2 s first-groups gap from #5287's catalog hold as that hold's cost. No follow-up quest.

Review: Codex gave a thumbs-up on head b24c3ad with no findings. Check, Test, and Quest pass. Interop is not required here, and it is red on main for unrelated quests such as unroutable.

(Written by Claude Opus 5.5)