I have seen this error when

dask-worker

s that are spun up by the

DaskTaskExecutor

are either paused by dask when it detects low memory, or when a

dask-worker

is killed somehow -- this last one happened to me when a

sig bus

error was thrown by the compute node I was running on.

First of all thanks for your message

I thought so for my case as well -- but there was something about the number of workers and exactly how my code was scaling that made dask pause a worker. In any case, it is something to look out for.