As projects grow and age and new features are introduced, there may come a time when it's probably better to start anew. Some features may require so much rewriting that it would be easier (and maybe faster) to just rewrite the whole thing from scratch. Also, then you have the opportunity to apply what you've learned from the first project and thus improve the code and design of the successor.

Is there a way of recognizing when this time is? Is it just a gut feeling, or could you say that "after x years, you probably should start over", or some other rule of thumb?