ManagedServer during shutdown tries to send a "shutdown" op to the remote server, triggering a graceful suspend.
Problem is, the TransactionalProtocolClient it uses is using a ManagementChannelHandler that may have been shut down. The service that controls the lifecycle of the ManagementChannelHandler (i.e. ManagementChannelOpenListenerService) is written to ensure that it isn't shutdown open until incoming requests complete, but there is nothing preventing shutdown when client uses are desired.
The specific area where this can be a problem is the assert !shutdown at
One possible solution is to distinguish locally initiated active operations from remote ones. AIUI the goal is to reject further remote ones.