It still seems to be cancelled for this PR, so I need to investigate further.

It seems this is not related to the timeout minutes, it just keeps happening randomly. It can even happen when the job queue is not that full but most of the times it happens when this action is waiting in the queue for some minutes (although it happened sometimes even only after some tens of seconds).

I tried different debugging attempts but I could not figure out a conclusion. The things that I tried:

• Playing with the timeout-minutes for the only job but this does not seem to have any effect because the job is cancelled before it has even started.
• Adding the DEBUG flag to get some debug logs from rokroskar/workflow-run-cleanup-action but this does not work because the workflow is never even started and it is cancelled already before it reaches this step.
• Removing the synchronize type from the event types but then we won't have the job status available when new pushes happen for the PR + I noticed many times this is actually cancelled when the PR is opened. The assumption here was if it is somehow related to the fact that the "push" actions won't trigger the action and I wanted to test if it is somehow thinking it needs to cancel this job because push actions shouldn't trigger it (as this is the only difference in this regard with the other actions).
• Looking through the workflow runs from the GitHub API if there was any extra information available there but there wasn't anything relevant. The cancelled runs just have "status": "completed" and "conclusion": "cancelled" in the API.
• Tried to search if some other people were experiencing this particular issue but the only answer I found eventually was just "it must have been a temporary issue with GitHub". There were some people running their own workers saying that there are some cases where the workflow run is cancelled when it fails to either find a worker for the job or when the worker creation fails for one reason or another. But there's not much information available about these for us (and no logs about the cancelled run), so I guess we just need to keep on guessing. It would also seem odd that this doesn't happen with any of the other actions.

Therefore, I will close this one with the following notes:

• This is likely not related to the action's configuration
• This can be some internal problem with GitHub but we have no information available about that
• If someone wants to continue from this in the future, I guess the next step would be to ask any information from GitHub support (and link the particular workflow runs that were automatically cancelled, I won't link any here because they will disappear at some point)