Z2JH singleuser pods not surviving hub outage?

In what circumstances is Z2JH supposed to terminate singleuser pods?

I recently had a failed chart configuration change, during which all existing singleuser pods appear to have terminated (whilst the hub itself was crashing every few minutes). However in the past we were used to seeing the singleuser pods survive even if the hub was restarted or temporarily unavailable. From a user perspective, having an active session terminated (losing unsaved memory contents) is far worse than merely being unable to interact with it temporarily (while waiting for a failed hub config change to get reverted by admins). Has there been any behaviour change, or is there any documentation of the circumstances in which singleuser sessions should survive a hub outage?

Thanks!

A user pod in a z2jh setup is terminated by jupyterhub only for two reasons i think: jupyterhub does it through a request to stop it, or jupyterhub does it through a request to stop the pod from the jupyterhub-idle-culler. Hmmmm… May be another reason, like a health check during jupyterhub startup? I don’t remember so well.

Were they terminated by jupyterhub at all? Then you should see a log entry about it i think.

If you saw terminations, maybe it was because they failed to report activity to jupyterhub, because of some networking issue?