Github scheduling jobs on completed runner pods

[kwngo]

We're using controller v0.22.3 and we're having an issue where pr checks are being scheduled on runners whose underlying pod is in Completed status in kubernetes. When controller restarts the pod, the pr checks seem to skip and disconnect.

Has anyone else seen this issue?

[kwngo]

Our current hypothesis is this is also related to our issue with runner de-registration 😅 #1544

[kwngo]

Can confirm upgrading to latest runner and using --ephemeral flag has fixed this issue.