Create a correct non-crashing CR with 2 broker instances Wait for them to spawn Modify CR to include something which would crash brokers (in this instance, that would be jolokiaEnabled variable as this is currently bugged; however anything like a non-valid init container URL would also work) Wait for updates What actually happens is, at this point 1 broker of the 2 starts restarting, reaching a "crashloop" state according to Openshift status. This crashloop can not be reconciled from, reverting the change to Broker CR does nothing as Operator doesn't try to update the statefulset from this state.   Only way to recover from this situation is SS removal for operator to re-create.