This is the same class of problem as
On startup, we start a "timestamp monitor" on the StorageEngine, which is responsible for completing the second phase of the two-phase drop we employ to drop collections and indexes.
After completing a FCBIS, we do not start the timestamp monitor again after re-constructing the storage engine. As a result, we will never actually drop any collections or indexes from the storage engine. This can be fixed permanently by restarting the server.