Add fail-fast feature to make it more kubernetes-friendly
At the moment whenever delivery gets "stuck" as per gitlab-org/charts/gitlab#2576 (closed) .
To fix that we could add configuration for delivery operators timeout and fail entire mail_room if we get past timeout. Kubernetes will respawn the Pod. Alternatively we could create "flag" file on disk (pre-configured as well) and liveness check in Kubernetes could check for it's existence and trigger re-spawn of Pod as well.
Not sure whether this change would sit with this fork or upstream.
Edited by Dmitry Makovey