Recently we discovered a memory leak during planning where a partial plan is built for a subtree of the query, then we later find that we cannot index another branch of the query and discard the larger plan in a way that leaks the partial plan.
This is known to be fixed by SERVER-35455 so this is simply adding a regression test that we can use to verify the fix on older branches while backporting SERVER-35455.
- related to
- 
                    SERVER-35455 QueryPlannerAccess should hold owned pointers by unique_ptr rather than raw pointer -         
- Closed
 
-