We currently don't exercise random combination of cleanup/compact and other ddls. 
For SERVER-122159, in order to fix a deadlock, both the operation have been rewritten to release and re-acquire locks at need, especially before the process logic that runs within a multi-doc transaction. 
We should exercise random or deterministic combination of cleanup/compact and other ddls, especially `dropDatatabase` and `renameCollection` to verify the robustness of the solution.