[3princip]

I agree with your points with a couple of caveats. I've written a lot of terrible code. For the good stuff it's usually not the existing lines of code or structure that are butchered rather the additions which do not follow the spirit of the original code. That ranges from trivialities such as code style of another developer or more serious issues of making a mess to fit a completely orthogonal new requirement, not utilizing existing functionality rather just doing similar things in different ways etc. In such cases a small refactor would suffice but that almost never happens due to time constraints.

On the topic of overly complex stuff not being touched, I find that's usually because no one understands it and hence others refrain from touching it lest it break.