ClusterPanicEvent is not delivered to cluster nodes

In a split brain scenario, the cluster safety job shuts down the host node and sends ClusterPanicEvent to the rest of the nodes of the cluster, to bring the whole cluster down. Because of a bug that was introduced recently, ClusterPanicEvent wasn't delivered to all nodes of the cluster, so they kept working, until the cluster safety job ran on each node.