We expect a node to leave the cluster as part of its shutdown sequence, but if the shutdown takes a long time then today the ClusterFormationFailureHandler continues to log a warning every 10s. We shouldn't be logging those warnings, it's normal not to be in a cluster any more at this point in the lifecycle of a node.

Relates https://discuss.elastic.co/t/master-not-discovered-or-elected-yet-when-stopping-elasticsearch-service/353377