I posted recently about Mega-Prizes, a term that Peter Diamandis used in a recent talk on large prizes for advances that seem impossible or almost so today, and that is also used in some academic prize literature for advances that "trigger radical new directions in science".
Diamandis presents the concept in an XPF blog video.
I guess if I were allocating $100M in prize money (the lower end of what Diamandis is suggesting), I'd probably allocate it for 5-20 smaller prize competitions, since the method has been shown to work well compared to other alternatives in that range. It would be too bad if the $100M opportunity for progress were blown because the challenge was, in fact, as impossible as it seemed.
I guess I'm more inclined to go for several smaller goals that give incremental progress than going all-out on a big project:
Mars Observer, or several smaller Mars robotic missions?
For a more difficult choice, Cassini, or 6-8 Discovery missions?
ESAS/Constellation, or Falcon 1, Falcon 9, Rocketplane, Space Ship 2, Lynx, Spaceloft, Pixel, JP Aerospace, Taurus II, XA 2.0, and numerous other smaller space access projects that could happen with the same funding?
Of course I'd also be inclined to do the mega-prize if the concept generally made sense and that's what the prize funder wanted ... it could work!