It only seems to work when the person doing the work has good judgement about what things need to be right and the manager trusts them. If there are communication and trust issues, then there will be the constant back and forth between "what makes us money" and "what needs to be fixed."

IMO, good judgement about the things that need to be right enhances the ability to make money. Too often I see manager-type people disregard the technical grievances of a team as ambiguous/unimportant things, while at the same time wondering why nobody can deliver and get things done. There is very often a direct connection!

I think most of us want our code to see the light of day. If given an actual problem to solve, we aren't going to toil away at tiny details forever. We want to see it run!

A good manager knows this and will give you room to find that balance yourself, unless you have demonstrated you need help with that.

I read an interesting twist on perfectionism: try to the "perfect compromise". Quality has multiple dimensions and you can't optimize all of them perfectly. Instead, choose which dimension is actually most important for the current problem and prioritize it.

A few years ago I invented[1] the term 'acceptimal', meaning 'acceptably optimal'. It's not the absolute best solution, but it's a pragmatic compromise between time, cost, performance, and simplicity which performs acceptably close to optimally.

[1] Or at least I've never heard anyone else use it who didn't get it from me.