[kixelated]

Summary

Plans the follow-ups the maintainer picked on 2026-11-02 from the PRs in that day's /quest-complete round. Each quest links the PR it came from.

New quests

• m0, flakes: moq-c's gated callback test frees nothing early [XS], from the SIGABRT on moq-dev/moq#7461.
• m0, CI: Binary release publishers run on their tag [S] (requires publish-release-only, moq-dev/moq#7431) and The flake lane skips hosts that can't map user namespaces [XS].
• m1: moq-crypto replaces moq-sign and moq-e2ee [M]; A resume more than one group behind never hangs [S], from moq-dev/moq#7631; moq-quic's idle connections keep no burst-sized state [S], from moq-dev/moq#7528; Stats name the route instance's epoch with a type [XS], from moq-dev/moq#7489.
• m2: Browsers expose WebTransport receive credit [XS, condition] and JS fastest-reader pacing [M], from moq-dev/moq#7624; fMP4 import at 1-byte chunks [S], from moq-dev/moq#7628; The relay bench reports io_uring pool size and RSS [XS], from moq-dev/moq#7233; MKV imports ffmpeg's multichannel FLAC [S], from moq-dev/moq#7629; Revisit the relay's default QUIC idle timeout [XS], from moq-dev/moq#7096; The media test checks browser feedback [S], from moq-dev/moq#7222.

Edits

• Interop coverage covers drafts 18, 20, and 22 (from moq-dev/moq#6972).
• hls-reinit-same-codec now requires the m0 declared-grid export, as its Plan asked once moq-dev/moq#7616 landed.
• Automatic recovery records that whether the in-progress group survives a resume depends on the upstream cache being warm or cold.
• Fastest-reader pacing moves JS parity to m2 and gains a relay-level serve_ws windows test.

Already covered, so not planned again: the moq-tokio, JS close-drain and boundary-sync flakes (m0 test-flakes-2), jitter on a same-codec reconfigure (hls-conformance), thin vs fat LTO (audit-compile-time), and DASH across a reconfigure (dash-reconfigure).

Decisions

• moq-c gated callback: ✅ m0 · m1 · m2 · drop
• Timestamp check in moq-sign and moq-e2ee (maintainer: combine the packages before the cut instead):
  • Name: ✅ moq-crypto · moq-secure · keep moq-e2ee and absorb signing
  • JS: ✅ merge @moq/sign and @moq/e2ee into @moq/crypto · Rust only
  • Layout: ✅ plain sign and e2ee modules · Cargo features
  • Published moq-e2ee 0.0.19: ✅ final deprecation release pointing at moq-crypto · leave as is · yank
• Relay serve_ws windows test: ✅ fold into fastest-reader (m1) · own m2 quest · drop
• Browser receive credit: ✅ m2 condition quest, with JS parity moved to an m2 quest that requires it · m1 required by fastest-reader · drop
• Resume more than one group behind: ✅ m1 · fold into hole markers · m2 · drop
• fMP4 1-byte-chunk slowdown: ✅ m2 · m1 release-perf · drop
• moq-quic readable events and idle bytes: ✅ m1 · m2 performance survey · drop
• Bench pool gauges and RSS: ✅ m2 · drop
• ffmpeg multichannel FLAC in MKV: ✅ m2 · m1 · drop
• DASH across a reconfigure: ✅ m2 (already planned as dash-reconfigure; only the hls-reinit Required edit is new) · m1 · drop
• Interop draft 20: ✅ add draft 20 · keep 18 and 22
• Route health resume and idle timeout: ✅ fold into route-health m2 · only the resume · drop both
• Feedback media test: ✅ m2 stats epic · drop
• Typed stats epoch: ✅ m1 · m2 · drop
• Binary release publishers on their tag: ✅ m0 ci · m2 · drop
• uid_map flake-lane skip: ✅ m0 ci · drop

Checks

quest check (551 documents) and quest gate nix develop --command just check origin/main pass.

Review

Codex gpt-6.1-sol, medium effort, via quest review --base origin/main, session 01a19e89-c9fb-71a3-9e22-a7559d79d2d2, round 1, reviewed head a663fb8e28978a45030470d6de11b177d0500940 against base 2738588cfc: no findings.

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: a663fb8

No actionable bugs found. The follow-ups target root causes, retain release and browser prerequisites, and favor existing mechanisms over new compatibility machinery. Separating investigation from a presumed fix is appropriate for the performance and route-health work.

Static source review only. I did not run the documentation gate, reproduce the reported failures or measurements, or verify browser capabilities.

[pullfrog]

ℹ️ No critical issues — two minor plan-text corrections inline.

Reviewed changes

• New m0 quests: release-tag-publishers and userns-uid-map (CI, from #7431/#7461), and moq-c-gated-callback (test flakes, the SIGABRT from #7461), indexed in their READMEs.
• New m1 quests: moq-crypto (merge moq-sign/moq-e2ee, from #7614), ietf-resume-behind (#7631), quic-readable-dedupe (#7528), stats-epoch-type (#7489).
• New m2 quests: browser-receive-credit + js-fastest-reader (#7624), fmp4-byte-chunk-perf (#7628), bench-uring-pool (#7233), mkv-flac-channel-mask (#7629), route-health/idle-timeout (#7096), stats/feedback-media-test (#7222).
• Edits: interop-coverage adds draft 20, hls-reinit-same-codec gains its hls-media-grid Required, fastest-reader moves JS parity to m2, route-health/recovery records the warm/cold resume behavior, and the milestone/epic READMEs index the new files.

Every cited PR number, file:line, symbol, and cross-link checked against head (including on_track, gated_callback, TrackState::answered, receive_bursts, moq export echo, hls-media-grid, and the 0.0.19 crates.io release) resolves correctly; no duplicate quests found.

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using deepseek-v4.1-flash (free via Pullfrog for OSS) ｜ 𝕏}

[pullfrog]

This note tells the reader to substitute "18, 20, and 22" manually, but line 42 still opens Drafts 18 and 22 only, nightly: and line 43 links moq-transport 18 and 22 only (whose title is now "moq-transport 18, 20, and 22 only"). AGENTS asks stale text be fixed inline, so editing those two spots removes both the contradiction and the override note.

[pullfrog]

GroupFrame::epoch is typed moq_net::Instance (rs/moq-stats/src/lib.rs:135), not a bare integer; #7489's suggestion was that it reuses moq_net::Instance for its UUIDv7 text and would read clearer with a type of its own. Rewording to "rather than reusing moq_net::Instance" makes the Goal match the actual starting point.