It currently says:
"Operations on a document must be atomic and ordered, so operations on the same namespace will be put on the same thread to be serialized."

The second part is not entirely correct as we will still parallelize ops on the same collection if they are touching distinct documents. See here.