When there is an error on HTTP listen, webhook does not handle the error in a way that recovery is possible and instead hangs without printing anything useful on the logs.

Seen after this change https://issues.redhat.com//browse/OCPBUGS-20104 where the webhook was re-configured to run as non-root but listen would fail on upgrade as the old webhook instance was running as root which causes an error due to the SOREUSE socket option.

The webhook should crashloop instead which would provide a chance of recovery although the recovery itself might still be racey depending on whether k8s is able to kill the old webhook instance before noticing the crash of the new instance.