The test starts 2 nodes (A and B), kills A and starts C. Test cleanup stops C and gets stuck in DCM.undefineConfiguration as the method is iterating through caches and finds out that ___counter_configuration is not started yet, waiting for initial transfer.

The node waits for initial ST because by the time C starts the cache is degraded on B and initial transfer cannot continue.

Proper fix would be letting the cache start but operations throwing AvailabilityException, but workaround would be just keeping wired unstarted caches in another collection directly for the purpose of undefineConfiguration.