Cluster join_timeout and partition-induced formation failures in Neo4j causal clustering

Neo4j causal clustering configures the cluster.join_timeout parameter, defaulting to 60 seconds, to manage member discovery during formation. Network partitions that isolate nodes before quorum is reached generate log entries such as 'Cluster formation failed: timeout waiting for initial hosts', indicating unresolved membership. The conditions under which a subsequently healed partition triggers leader revalidation remain ambiguous in documented behavior. Clock drift across members and hostname obfuscation in containerized deployments further complicate log correlation.

Under what circumstances does a rejoining node invalidate an existing leader? Which log entries confirm that cluster.join_timeout expiration precluded formation despite later member availability? How does the election timeout interact with partition healing to determine final cluster state?