When a stream is deleted, its uuid is recorded in the volume's delete journal, to be tracked for a period of time (by default 2 weeks). As long as one of the stream's replicas is online at the time of the delete (otherwise, the application will receive a 404 error and should retry the delete later), it will notify other replicas that might have missed the delete request. Eventually, the uuid is removed from the delete journal along with all evidence the stream ever existed and was deleted. After that point, any nodes that have been offline for a very long time have the potential of resurrecting the deleted content. For this reason, nodes that are offline for longer than a specified period of time, called the RECOVER_TIMEOUT, cannot rejoin the cluster without reformatting their disk volumes and starting out as a new, empty node.