What the numbers show
queued ≥ 60s or ≥ 20% of the jobQueue time is the gap between a job being created and a runner picking it up. It comes from concurrency limits on your plan, busy self-hosted pools, or rare runner labels. It's only visible in run exports or the job's own timestamps.
The change
Cancel superseded runs so they free runners
concurrency: group: ci-${{ github.ref }} cancel-in-progress: trueVerify the verdict stays the same
- 01Compare created_at with started_at for the same job before and after.
- 02Cancelled runs are only ones a newer push replaced.
- 03The latest run's conclusion is unchanged.