One method I’ve found: realizing that something I don’t like may simply be ripped out in its entirety at some point, and therefore I may be wasting time obsessing over its minor details right now. (And deleting hundreds of lines of an awful module later is a much better feeling than any amount of optimization beforehand.)
Also, I found that I tend to perfect only the things that I work on alone. As soon as anyone else is involved, you have to start worrying about things like not making difficult merges, and most perfections fall into the category of “PITA merge”. That grand 50,000-line code-reformatting that someone did will not really help, it will just come up again and again across 4 branches that they didn’t know existed, as each person trips over the same things. And I guarantee there is at least one tiny bug in that reformat somewhere. So don’t do it; keep things as close to what they were before as possible, pretty or not, for as long as it lives, unless you can be absolutely sure that there won’t be other branches out there.