In test_backup18.py we check that incremental backup is no longer configured after a force_stop=true. This checks the live running system.

We should enhance that testing for simulating a crash immediately after that completes (closing the force backup cursor). Once the cursor is closed, make a copy of the database directory and open a backup cursor with any old IDs should fail. When the force_stop cursor close returns there should be no remnants of any earlier ID visible on restart if a failure happens.