Questions about clustering. Is there a legitimate use case for a cluster with only 2 nodes (say, 2 machines where one is the primary and one a failover)? It seems like there could never be "consensus" about which node had a problem in the event of an issue where they lose contact.
Also, it's my understanding that if a node gets kicked out of a cluster, it can never rejoin the cluster unless its ActorSystem is restarted so that it gets a different UID. It seems like this would make
Would make it harder for a node to recover from an intermittent failure and just "come back online". I was wondering if the behavior of requiring the disconnected node to completely restart before rejoining is more of a workaround for some technical limitation, or if there is a design reason that made this a superior approach?