I know this is not formally an ordered list, but it looks cart-before-the-horse-ish to have catching future regressions before verifying that it works in the first place. (This comment is effectively a reply to the OC's claim that we don't test to verify.)