DHH actually did follow Spolsky for 7 years and only after running out of ideas and constantly hitting brick walls they started investigating a rewrite to rejuvenate the product based on new ideas that weren't around when they first built it. I guess, the answer is, rebuild only if you can't build your new ideas anymore and you can't attract new customers. (He calls it becoming japan in the talk.)

I never rebuild/rewrite an entire product - ever - it's bad and Netscape as well as others failed because of it. I try to refactor and evolve based on user feedback. A button no one ever uses doesn't need to work.