If a realfast_controller instance is redefined, then some commands suggest that there are multiple distributed clients going.
In [38]: rfc.restart()
2018-12-06 21:14:16 INFO distributed.client Receive restart signal from scheduler
2018-12-06 21:14:16 INFO distributed.client Receive restart signal from scheduler
2018-12-06 21:14:16 INFO distributed.client Receive restart signal from scheduler
This happens when event loop is not closed properly (with shutdown or close methods). Issue should have been fixed by https://github.com/dask/distributed/pull/1073, but that was a while ago.
If a realfast_controller instance is redefined, then some commands suggest that there are multiple distributed clients going.