Today $unionWith aggregation command is executed sequentially. EG first we query collection A and then collection B and then the union occurs.
The process should be parallelized so the query part will run in parallel while the union will be done as best effort tree merge try to speed up the overall Elapsed Time of the query