A cleanup transaction in our Neo4j database lead to a deadlock on one of the largest nodes, leading to a large transaction queue and exhaustion of available connection pool to the database. This caused instability and in some periods downtime, and required reboot of the affected database.
