we had a short discussion on a Github issue regarding keeping worker code separate from the main application code because my worker code involves some funky stuff and has heavy dependencies (CUDA, etc.)
If I may refine the question:
(1) Is it possible to keep worker code completely separate, i.e. in a separate repository and only ever run the worker code on remote machines?
(2) as a consequence, is it possible to ask the broker for the names of the workers that joined?