I think there's a balance to be struck between code quality and delivery speed.

GNU Hurd is a well known example where they spent a so much time on getting the code right that the project became irrelevant. And of course you made the case for erring on the "too much code too quickly" side