For update and delete, we have logic to translate the bucket namespace (e.g. sent from mongos) back to the view namespace when we set up the CurOp. But for insert, it looks like we always use the namespace from the received request. And I think this means that for writes sent from mongos to the shards, they will always be logged with the bucket namespace. I am not sure what's the desired behavior here.

This test confirms this mismatching behavior.