I've generally found that the best solution to a problem is also the simplest. If the complexity of your solution starts to grow more than linearly as you work

I've generally found that the best solution to a problem is also the simplest. If the complexity of your solution starts to grow more than linearly as you work on it, if it starts to look like the only way to finish it is through brute-force labour, then I've found that backtracking to right before things start to get complicated usually yields an "aha!" moment. A simple, elegant solution will be easier to implement and maintain, more fun and satisfying to work on, and will most likely even solve the problem better. You don't always have that luxury, of course - some problems don't necessarily have a simple solution. But for me, when things get complicated I take it as a sign that I've taken a wrong turn somewhere. I'm not sure if this actually answers your question. ;)