It's not that, so much as manual testers will /notice/ that code is slow, and ask questions about it, whereas automated tests will run the script no matter what

It's not that, so much as manual testers will /notice/ that code is slow, and ask questions about it, whereas automated tests will run the script no matter what (unless someone thought about it in advance) Likewise, you /can/ code an automated test to simulate sudden network loss, but a human will just randomly decide to try it, without someone pre-programming him