“Consider something like a failover for a database from a hard failure. While the recovery itself might be fully automated (and should be), triggering the test itself is likely best done manually. Additionally, it might be far more efficient to determine the success of the test manually, although developers should still encourage scripts and automation.”