[kixelated]

Problem

On the 0.17 line, a FETCH of a group that a live subscription delivered only in part, after that subscription went idle, returns the cached frames and then hangs forever (#5348). When the last reader leaves mid-group, the front parks a warm copy (warm_copy/warm_rebuild in model/origin.rs) that rebuilds the open live-edge group, and nothing writes to it again. resume::Consumer::cached_group served that open copy as a FETCH hit, and no upstream SUBSCRIBE or FETCH followed.

Seen downstream as a recorder restarted within the cache window permanently stalling a rendition.

Approach

Per quest/m1/release-warm-fetch.md on main:

• An unfinished group in a warm segment is a FETCH miss (cached_group takes a finished flag, set for warm segments).
• Fetching::poll_latch never latches a warm segment. The pending fetch holds demand, the front re-splices, and the FETCH goes upstream on the new copy.

Regression: rs/moq-net/tests/idle_cancel_refetch.rs drives it over a lite-06 mock session. It fails on release with group 1 hangs: Elapsed and passes with the fix.

Impact

• Public API: none. Wire: none.

Alternatives

• A spliced resume::Group continuing the open warm group: more code on a line main already deleted (fix(net)!: resume route changes by reading the routes' copies; a path is one broadcast #4741).
• Aborting open groups at park: breaks a returning subscriber continuing an open warm group (returning_reader_continues_an_open_warm_group).

Follow-ups

• The same test lands on main as a guard, retiring the quest: test(net): a group cut by an idle cancel is refetched whole #5366.

Closes #5348

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[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 2026-10-12T20:34:55.184840Z	9a50b9f	PR opened

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 9a50b9f against release (0393334).

No actionable bugs found. The two guards address both ways a FETCH could get stuck on the parked copy: cached incomplete groups miss, and upstream requests wait for a live replacement. Finished cache hits remain available, and the existing demand lifetime supplies the re-splice. This is a suitably narrow fix that preserves live subscribers' open-group continuation.

Inspected all three changed files and surrounding cache, demand, park/takeover, group-completion, and mock-session code. Verification was static through GitHub; I did not execute tests. CI was still running when checked, so this does not establish a test pass.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9a50b9f273

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".

[kixelated]

Merge summary:

• An unfinished group in a parked warm segment is now a FETCH miss, and Fetching::poll_latch never latches a warm segment, so the FETCH waits for the re-splice and goes upstream instead of stalling after the cached frames.
• Regression rs/moq-net/tests/idle_cancel_refetch.rs fails on release without the fix.
• OpenAI review: no findings. Codex P1 (move the test inline) resolved per the reply above: it is a full-session test over the shared mock harness, and release already keeps ~25 such files under rs/moq-net/tests/.
• No API or wire change. Merging into release with a merge commit; test(net): a group cut by an idle cancel is refetched whole #5366 lands the main guard and retires the quest.

(Written by Claude Opus 5.5)