test(moq-cli): the TS passthrough jump test holds under load without pacing

[kixelated]

Quest: quest/m1/test-flakes-2/ts-passthrough-jump.md (deleted here, done).

Fixes moq-cli's ts_passthrough_crosses_a_relay_through_a_flagged_jump, which failed under load on main. The 15 ms wall-clock pacing is gone: the second copy now decodes all at once, and a run takes about 0.6 s instead of about 4 s.

Root cause

Two causes. The relay one landed on main first, in #4225; this PR now carries only the test-side fix.

1. Relay (fixed by fix(net): keep a lite subscription's demand until its groups drain #4225). A lite publisher dropped its track::Subscriber once its subscription reached the track's end, while its group streams were still draining. On a relay that cancelled the upstream subscription and stranded an older group's tail. fix(net): keep a lite subscription's demand until its groups drain #4225 keeps the subscriber in Drain until the groups finish. Its regression finite_relay::relay_keeps_groups_in_flight_past_the_end runs on simulated time and fails without that fix (moq-lite-05: "the track never ended"). This PR first fixed the same bug by removing Drain; after merging main it keeps fix(net): keep a lite subscription's demand until its groups drain #4225's version and drops its own relay_end_in_flight test, which covered the same scenario.
2. Test (fixed here). On moq-transport-14 the recording joins the publisher's current group when its SUBSCRIBE reaches the publisher through the relay, not when it is sent. Under load that landed after the second copy's first group, so the recording missed the end of the first copy and "both copies crossed" failed. The test now subscribes and takes the recording's first group before decoding the second copy.

Decisions

• Pacing: fix the relay bug and drop the pacing. ✅ Chosen by the maintainer.
  • Land with the pacing and plan a follow-up quest.
  • Replace the pacing with event pacing.
• Overlap with fix(net): keep a lite subscription's demand until its groups drain #4225 (sub-agent pick, flagged for review): ✅ keep fix(net): keep a lite subscription's demand until its groups drain #4225's Drain fix as merged and trim this PR to the test fix.
  • Replace it with this PR's simpler version (no Drain state, an ended flag on TrackRun). Smaller, but it rewrites code that was just reviewed and landed.
• The quest's Required on the moq-time crate, added on main after this PR was opened (sub-agent pick): ✅ delete the quest anyway. With the pacing gone, the test has no wall-clock dependency left beyond its failure deadlines, so it no longer needs a controlled clock to hold under load.
• Fail loud when a group is stranded anyway: left to SUBSCRIBE_DROP, whose Plan already owns it.

Test

• just check: passes.
• just test interop --all, run twice: every cell passes except the three tail rust -> * lanes, which fail the same way on main. Since feat(net)!: an undeclared Rust timescale means untimed #5083, create_track(name, None) is untimed, but test/interop/clients/rust/tail.rs still writes Timestamp::ZERO ("frame timestamp doesn't match track timescale"). That is a follow-up, outside this PR. On the first run, rust -> go, rust -> c, and one relay idle-out also failed; none repeated on the second run.
• The CLI test under load (8 parallel loops pinned to 2 cores next to 8 busy loops): 64/64 pass with this PR. main's paced version fails 24/24 under the same load, every time on "moq-transport-14: both copies crossed".

Public API and wire

Public API: none. Wire: none. Test and quest changes only.

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[kixelated]

Outcome: the race is fixed at its cause and the PR stays a draft. The 15 ms wall-clock pacing is still there because removing it exposes a relay bug (see the description).

Open decision for the maintainer: what happens to the pacing.

1. Land this as is and plan a moq-net quest to fix the relay dropping its upstream subscription while groups are still in flight. That quest drops the pacing as its regression test. (Recommended)
2. Fix the relay bug in this PR. That widens a test-only PR into moq-net.
3. Replace the pacing with event pacing (the recording reports each group, and decode waits on it). That works around the same bug and adds plumbing the relay fix would make unnecessary.

(Written by Claude Opus 5.5)

[kixelated]

Iterated per the maintainer's choice of option 2: the relay bug is fixed in this PR and the 15 ms pacing is gone.

Root cause: a lite publisher dropped its track subscription as soon as it reached the end, before draining its in-flight group streams. On a relay that cancelled the upstream subscription with an older group's tail still coming. The new relay_end_in_flight regression fails on moq-lite 05/06/07-wip without the fix and passes on every version with it.

Left to quest/m1/subscribe-drop.md: a group stranded some other way still hangs rather than erroring, because resume waits on a dead route. That quest's Plan already covers it.

PR remains a draft. CI's Check and Quest jobs are red only from main's dangling quest links, which #5177 fixes.

(Written by Claude Opus 5.5)

[kixelated]

Heads-up on overlap: #4225 fixes the same relay publisher drain cause, but keeps the track subscriber inside Drain until its group streams finish, where this PR removes the Drain state. #4225 is landing first, so this PR needs to rebase onto it and reconcile (or trim to its moq-cli test changes) when it resumes.

(Written by Claude Opus 5.5)

[kixelated]

@codex review

(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 2026-10-12T15:34:54.667563Z	90edb78	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.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Keep it up!

Reviewed commit: 90edb78fd3

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[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 29 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: 20bf85aa-1582-4935-89f2-d092f8b3f2d8

---

📥 Commits

Reviewing files that changed from the base of the PR and between 9050086 and 90edb78.

---

📒 Files selected for processing (3)

• quest/m1/test-flakes-2/README.md
• quest/m1/test-flakes-2/ts-passthrough-jump.md
• rs/moq-cli/src/publish.rs

---

• 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.}