Ah; but how annoying it is to discover something like the inverted retina bug, only to figure out it is effectively unsolvable now due to all the follow-up architecture decisions built on it?

Yes, don't give a crap about doing it right so all your costs instead move into support and maintenance (which in this analogy is healthcare)