Gaetan and myself noticed that sometimes stopping the server takes a long time; looking at java stack gives only vague indications about the shutdown process.
The new disk space monitor register itself as a listener for shutdown, and right after deregistering all disk monitors deregister itself.
This is done before the MonitorProvider service itself has been shutdown, so there may be a lingering thread still scheduling the run() method.
Do not deregister itself so that the shutdown process can cleanly stop everything.