ci-status: a stale ci-lanes failure ends the carry-forward wait while its full run is re-running

[kyle-sexton]

Summary

In carry-forward mode, ci-status ends its wait as soon as the newest ci-lanes status is success, failure or error. It does not check whether the full run that wrote that status is being re-run. After a full run fails and someone re-runs it, a contract-only event on the same SHA (a body edit, a label flip) reads the old failure at once and goes red. It never waits for the re-run, which is the run the wait exists for.

The failure message also sends people the wrong way. It always says re-run the full workflow. Once ci-lanes on the SHA is success, re-running the red contract-only run is enough, but the message never says so. A new commit (every lane again) is the natural reaction.

Evidence

Reproduced in melodic-software/claude-code-plugins#4670 with a throwaway pull request (claude-code-plugins#5581, closed). Consumer pin 4610c31e (v0.27.1). run.sh on main differs from the pin only in comments. Head SHA ad99ed0ce859b0fa1071d48f30c410528ee331f8, all times 2026-09-30 UTC.

Time	Event
05:56:46	Full run 36673988526, attempt 1, fails. ci-lanes=failure recorded (status 55241522644, target_url .../actions/runs/36673988526).
05:57:10	Attempt 2 of 36673988526 starts (run_started_at).
05:57:16	A body edit creates contract-only run 36675804726.
06:00:01	Its ci-status logs ::error::no successful ci-lanes status on ad99ed0c...; re-run the full workflow with no Waiting line, while attempt 2 is still in flight.
06:12:49	Attempt 2 passes and records ci-lanes=success (status 55242254969).

Re-running 36675804726 after that logged Carried forward: ci-lanes is success and replaced the stale check run in the same check suite. So re-running the contract-only run is the right remedy once ci-lanes is success.

Proposed fix

1. A settled failure or error ends the wait only when no full-run re-run whose current attempt started after that status's created_at is in flight. A success still ends it at once.
2. When the verdict is not success, the message keeps re-run the full workflow and also says to re-run this run once ci-lanes on the SHA is success.

Constraints:

• Do not bring back mutual waiting between contract-only siblings. Reading the settled status each poll is what releases them today.
• Keep every path fail-closed. A timeout, a missing status or a skipped lane never passes.

Related

• fix(ci): contract-only ci-status reads a stale failure during a re-run claude-code-plugins#4670 (consumer tracker, reproduction results)
• feat(ci-status): bounded carry-forward wait and the branched contract-only concurrency group #557, fix(ci-status): order the carry-forward wait by run id so same-second siblings are seen #559, ci-status: carry-forward wait misses a same-second full sibling with a higher run id #561, fix(ci-status): wait on any in-flight sibling before reading ci-lanes #562 (earlier carry-forward wait fixes; none covers this)