Handle fed jobs when cluster is removed from fed

while the cluster is down. The cluster will figure out what changed
after starting up or after resuming from using the cache.