Instead of using our own task pool, we should leverage Dask distributed, as this will allow us to better consume resources from existing clusters.