You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(ci): two rate-limit messages send sessions the wrong way #16653
Two CI messages about rate-limit holds misstate what happened, and sessions planned around the wrong cause on 2026-11-16.
Rate guard, fallback hold. When GitHub's refusal carries no reset header, the rate guard's error prints "holds until <refusal + 1 h>". That time is the 3600 s primaryFallbackSeconds (.github/actions/log-api-budget/rate-guard.mjs:50, chosen at :470-471). It is local to the process and never written to the stop file (headerStop, :495), so a re-run works as soon as the real reset passes. Two sessions read it as a real hold and planned re-runs about an hour late.
CI Status, uncounted review. CI Status says a lane "ran in summary mode or without a hold label" (.github/workflows/pr-require-checks.yml:7793) when the real cause was the lane's leg holding on a rate-limit 403 ("this leg holds and runs no review"). A session suspected pull request size because of it.
Evidence: pull request #16625's review-subscription run 43959008627 (2026-11-16 21:26Z), and CI Status runs 43938977243 and 43948536399.
Done when
A fallback hold's message says the time is a fallback because the refusal carried no reset header, or states the real reset read from /rate_limit.
When the lane's job recorded its rate-limit hold notice, CI Status names the held leg and that cause, not summary mode or a missing hold label.
Size: Small
Complexity: Low
Uncertainty: Known
Flags: none
Blocked by: none
Understood: two CI messages misstate rate-limit holds. (1) The rate guard's fallback hold prints a refusal-plus-one-hour time that is only a local fallback, not a real hold. (2) CI Status blames summary mode or a missing hold label when the lane's leg actually held on a rate-limit 403. The done-when is exact: label the fallback time or show the real reset, name the held leg and cause in CI Status, and add tests for both messages. Nothing is missing.
Related but not a duplicate: #16642 (the workflow token reaching its rate-limit floor). No merged pull request in the recent list fixes this. #16650 is out of scope, as the issue says.
The work-resolve-issue lane stopped on this issue (patch-protected-path): the change needs a path the issue lane may not push, which a maintainer session builds by hand.
It added host-run and set Work status to confirmed: an agent run on a host builds it, and it waits on no decision from a person.
In the run's words:
The run stopped early (patch-protected-path): its change needs a path the issue lane may not push, which a person changes instead. In the run's words: patch-protected-path: .github/workflows/pr-require-checks.yml. Done-when item 2 (CI Status names the held leg and the rate-limit cause instead of 'ran in summary mode or without a hold label') needs this file changed: the message is an inline echo at .github/workflows/pr-require-checks.yml:7793, and /.github/workflows/** is a path this lane cannot push. Item 1 (.github/actions/log-api-budget/rate-guard.mjs) and item 3 (tests for both messages) are in the same issue and change, so no partial patch was made.
What the run says it did not do:
Fallback-hold message in .github/actions/log-api-budget/rate-guard.mjs: not started, because the issue's change also needs the protected workflow file.
CI Status held-leg message in .github/workflows/pr-require-checks.yml:7793: protected path, not editable by this lane.
Tests for both messages: not written, because the run stopped before any edit.
Repository checks (check-changelog-fragments, check-changed-skills, check-affected) and tidy: not run, because nothing changed in the working tree.
Activity
github-actions commentedon Nov 16, 2026
Size: Small
Complexity: Low
Uncertainty: Known
Flags: none
Blocked by: none
Understood: two CI messages misstate rate-limit holds. (1) The rate guard's fallback hold prints a refusal-plus-one-hour time that is only a local fallback, not a real hold. (2) CI Status blames summary mode or a missing hold label when the lane's leg actually held on a rate-limit 403. The done-when is exact: label the fallback time or show the real reset, name the held leg and cause in CI Status, and add tests for both messages. Nothing is missing.
Related but not a duplicate: #16642 (the workflow token reaching its rate-limit floor). No merged pull request in the recent list fixes this. #16650 is out of scope, as the issue says.
Automated triage. A maintainer reviews it.
melodic-automation-lanes commentedon Nov 16, 2026
{"v":1,"actor_id":337595936,"title_sha256":"f20fee1782ecfc6ee8d07755a55af39d94ac502ab23f8dca941ff7dd9f3fee1d","body_sha256":"012a49b33f3a596d2dbcf3222ad86205421c12db67df45de8538f5c1ce20622e"}
github-actions commentedon Nov 16, 2026
github-actions commentedon Nov 16, 2026
melodic-automation-lanes commentedon Nov 16, 2026
The work-resolve-issue lane stopped on this issue (
patch-protected-path): the change needs a path the issue lane may not push, which a maintainer session builds by hand.It added
host-runand set Work status to confirmed: an agent run on a host builds it, and it waits on no decision from a person.In the run's words:
What the run says it did not do:
Run: https://github.com/melodic-software/claude-code-plugins/actions/runs/43971972300
To hand it back to the lane, remove
host-runand set Work status to agent-ready again.