Yup, if we want to beat the analogy further, I would say the constant attention to code "smells" and such is like cleaning behind the refrigerator daily when you see dust has landed. I think there is a great middle ground, where you actually consider the possibility that the improvement to the code will be a net loss (time investment to improve code and time it saves in the future). Lots of books about this, many people don't get the full theories, and instead take the talking points (x lines of code per method, don't do Y, do Z, etc).