It always seemed to me like the academic-style sabbatical was a nice solution to this. Instead of trying to convince a bunch of people that what you're going to do is going to be worthwhile, you develop a culture where capable people are regularly given time to work on whatever they think is important.

Clojure's a nice example: it's been hugely successful, but I have a hard time imagining anyone raising two years' worth of salary up front to work on their idea for an idiosyncratic Lisp implementation. It looks like he had to fund that work himself; how much more innovation would you see if this kind of extended project was common?