The FlowControlRefresher periodic task is shut down as part of the node shut down procedure. If there are insufficient tickets available, certain operations will not receive a ticket and will hang.

A potential fix for this issue could be to identify all operations that occur during shutdown and exclude them from flow control.
Alternatively, the PeriodicRunner shutdown block can be moved to after the final global lock acquisition during shutdown.
A third fix is to add synchronization handling upon shutdown of the FlowControlRefresher thread to circumvent ticket acquisitions. This third fix seems the most promising.