That's much less convincing for the FTS in particular, which needs to be reliable even (or especially) in early testing. Since you expect everything else to go wrong, you need to be able to stop it from going wrong on top of people.