In a clustered configuration, it is important that all CloudBees Flow servers that are set up to participate in the cluster can communicate with ZooKeeper and with each other though JGroups. So each CloudBees Flow server runs a periodic critical services health check to test that it has a valid session with ZooKeeper and that the server is part of the JGroups cluster for CloudBees Flow.
The health check is run every minute by default. If it fails after five repeated attempts, the server goes into bootstrap mode so that it cannot serve any requests. Once in bootstrap mode, the server will periodically attempt to reinitialize its services. If it is successful and is able to join the JGroups cluster, it can serve incoming requests again.