Hmm, interesting... so basically DuckDB works in this case because there's no way to parallelize the migration of a single volume anyway?

This is definitely a pretty niche case, though, so there must be something more general that this was built to do.