With this PR, dataset combine is no longer hardcoded to use a single-threaded dask scheduler. Hardcoding this, leads to regular UserWarnings as many/most users already have an active distributed client. I'm happy to be told that this is a bad idea, or that there's a better way to fix the issue reported in #596, but I'm submitting this now to get the ball rolling.
Related issue number
Closes #596
Checklist
[x] Unit tests for the changes exist (not really appropriate here)
[x] Tests pass on CI
[x] Documentation reflects the changes where applicable
Change Summary
With this PR, dataset combine is no longer hardcoded to use a single-threaded dask scheduler. Hardcoding this, leads to regular UserWarnings as many/most users already have an active distributed client. I'm happy to be told that this is a bad idea, or that there's a better way to fix the issue reported in #596, but I'm submitting this now to get the ball rolling.
Related issue number
Closes #596
Checklist