We should probably use dask instead of rolling our own scheduler.
[ ] Also, handle shutdown with Ctrl-C more gracefully. Currently, things randomly crash, interrupts get caught, … it's a bit unclear why this happens actually.
This would also allow us to actually parallelize things such as orbit closure search. Also, we could then get rid of hacks such as the remote progress queue and properly register it as a named queue with dask.
We should probably use dask instead of rolling our own scheduler.
This would also allow us to actually parallelize things such as orbit closure search. Also, we could then get rid of hacks such as the remote progress queue and properly register it as a named queue with dask.