Just a note, I've had no issues running pynisher inside of a subprocess running on a dask worker. This is likely not the issue. However there is some odd interplay between a `Nanny` worker class and a default `Worker` due to the fact the `Nanny` is in charge of rebooting failed workers. I don't really have more in depth info than that
Pynisher and dask parallelization does not work together.
Originally posted by @eddiebergman in #1023 (comment)
If we get it to work, revert #1023.