nose-randomly has the feature that nosetests path.One path.Two still runs path.One and path.Two in order, allowing reproduction with --randomly-seed=X much easier - once you've found a bad test order, just copy/paste it and rerun deleting paths one at a time.
pytest-randomly does not have this property because the pytest collection hooks are more generic. If possible adding this property would make it easier to debug such failures.
As noted by @tolomea