[TRI-1926] Auto-yielding a canceled run causes it to resume again
matt-aitken opened this issue · comments
To reproduce
- Have a run with several tasks in a row that are just a bit shorter than the serverless timeout.
- This will cause the run to "auto-yield" to avoid cutting a task that's halfway complete.
- Cancel the run during a task
The run looks canceled but when the auto-yielding happens it doesn't check that the run is canceled so it causes the run to go back to executing again…