Ended deployment runs
Read what happens to the GitHub workflows a deployment run dispatched once the run ends.
Ending a run cancels its workflows
When a run ends failed or cancelled — operator terminalization, a failed stage's status write, a rejected stage approval, or takeover by a newer run — every GitHub workflow it dispatched through a correlated dispatch intent (github_workflow_dispatch_intents rows whose request_id starts deploy:<project>:<run>:) is asked to cancel, newest attempt per stage, strictly after the close commits (deployment_run_workflow_cancellation). The cancellation is ordinary, so each workflow's if: cancelled() steps still run and can release what it started. A miss never reopens the run: yoke deployment-runs terminalize prints each outcome (cancel_requested, already_completed, run_unidentified with the correlation id to find it by, or cancel_failed with its recovery yoke github-actions run cancel REPO RUN-ID --project P), and every path logs misses with the same recovery. A one-shot dispatch records no intent, so it is not cancelled here.