The case
Ruleset main (read via API in the review): the required checks are the three OS legs, format and codecov/patch. They do not cover pack, validate release config and traceability. A PR that drops a README from the package or introduces a new untraced Must can merge green and only fails in the release run.
The merge queue is not enabled: zero merge_group runs, no merge_queue rule. #106 was closed on 2026-09-26 by PR #160 ("name merge-queue refs as 1.2.4-queue.N"); the activation itself did not happen (#165 asked for it and is closed too). Four places describe or simulate a case that never occurs: CLAUDE.md:36-44 ("#106 ... will carry the change"), ci.yml:7-17, GitVersion.yml:45-69 and eng/verify-merge-queue-version.sh (runs in every version job).
codecov/patch as a required check gates on presence, not on value (codecov.yml:18-20 is informational); if the upload is missing, the PR hangs.
Provenance correction: the merges of #81 and #83 are 2 min 31 s apart (9cd969b 22:50:19, 8c94d51 22:52:50), not "four minutes" (ci.yml:11-12, CLAUDE.md:37-38).
Confidence: high.
Why this needs a maintainer decision
Review section 7, item 11:
- Enable the merge queue (merge commit, not squash), or remove the configuration and the four places that describe it.
- Extend the required checks with
pack, validate release config, traceability?
- Take
codecov/patch out of the required checks, or make it a real gate?
Related: #106, #165, #85.
Deep review 2026-09-30, section 2.9 (B12); decision list in section 7, item 11; ordering in section 6, item 4.
🤖 Generated with Claude Code
The case
Ruleset
main(read via API in the review): the required checks are the three OS legs,formatandcodecov/patch. They do not coverpack,validate release configandtraceability. A PR that drops a README from the package or introduces a new untracedMustcan merge green and only fails in the release run.The merge queue is not enabled: zero
merge_groupruns, nomerge_queuerule. #106 was closed on 2026-09-26 by PR #160 ("name merge-queue refs as 1.2.4-queue.N"); the activation itself did not happen (#165 asked for it and is closed too). Four places describe or simulate a case that never occurs:CLAUDE.md:36-44("#106 ... will carry the change"),ci.yml:7-17,GitVersion.yml:45-69andeng/verify-merge-queue-version.sh(runs in everyversionjob).codecov/patchas a required check gates on presence, not on value (codecov.yml:18-20is informational); if the upload is missing, the PR hangs.Provenance correction: the merges of #81 and #83 are 2 min 31 s apart (
9cd969b22:50:19,8c94d5122:52:50), not "four minutes" (ci.yml:11-12,CLAUDE.md:37-38).Confidence: high.
Why this needs a maintainer decision
Review section 7, item 11:
pack,validate release config,traceability?codecov/patchout of the required checks, or make it a real gate?Related: #106, #165, #85.
Deep review 2026-09-30, section 2.9 (B12); decision list in section 7, item 11; ordering in section 6, item 4.
🤖 Generated with Claude Code