We have observed zoekt-webserver being killed by the watchdog. This typically
can happen on boot were a few other services are also competing for
resources. Used a failure count rather than a longer timeout since that is
usually how I configure the health checks in Kubernetes.
