All money issues aside, 4 years is also around the time you tend to get really sick of working on the same problem with the same codebase. And that's an outside bound, when the problem and codebase you were working on were both good ones. The wall gets hit even sooner if you're working on something less interesting.
All IMO and IME, of course.