@alex I think that would actually work, since the max parallelism would probably be hardcoded.