How does retire work?

If Node1 calls out a retire, all other nodes go into Fast Volume Recovery to see if they have replicas of streams that were known to also be on Node1 and make replicas.  At the same time, Node1 goes through each stream and does a rep query.  If the stream does not have two other reps on the cluster, it makes a new replica somewhere else, then trims it locally.  Otherwise, it just trims the local copy.