I like Apple's method: the 10 to 3 to 1 design winnowing/crossfertilization seems to end up with great products. (If only they'd done that with iTunes).
Porting that method to open source is a little problematical. You need a great curator/BDFL.
Problematically for those seeking 100% (which I tend to do because 100% perfect solution is like geek-heroine), the 80% products hit the market first and tend to make more money. :(
Olin Shivers, my most bookmarked and least read writer.