A VM was created on the cloud provider but never booted properly (sshd.service not found in serial logs — likely a transient image/cloud-init issue, since subsequent runners on the same type worked fine). The tray was in Creating state until the stale handler caught it
Issue is single 15-min threshold for any non-Running status. A Creating tray that never transitions to Registering within a couple of minutes is almost certainly broken — there's no reason to wait 15 min. Meanwhile Registered trays may legitimately idle that long waiting for a job
Possible solution is to have configurable per-status stale thresholds
A VM was created on the cloud provider but never booted properly (
sshd.service not foundin serial logs — likely a transient image/cloud-init issue, since subsequent runners on the same type worked fine). The tray was in Creating state until the stale handler caught itIssue is single 15-min threshold for any non-
Runningstatus. ACreatingtray that never transitions toRegisteringwithin a couple of minutes is almost certainly broken — there's no reason to wait 15 min. MeanwhileRegisteredtrays may legitimately idle that long waiting for a jobPossible solution is to have configurable per-status stale thresholds