when enough threads try to do it on the same database at the same time.
This is because all of them will try to grab the dist lock on the database. The dist lock algorithm doesn't have fair scheduling so it is possible for a single thread to be starved by other threads if there is enough contention on it.
- related to
-
SERVER-33639 Concurrent writes against non-existent database can fail due to distlock acquisition timeout at `createDatabase` time
- Closed