I haven't created any CG contest, but I'm pretty sure it's not that easy to create a viable, balanced and attractive game for everyone. Most notably, the ranking system seems a significant restriction to me, not allowing a more experimental approach aimed at exploring new kind of games.
I've particularly in mind cooperative as well as open-ended challenges, with not real victory conditions (at least, not calculated the way it is done now). More notable changes, such as getting rid of the turn-based sequence also come to my mind, but the actual template is already sufficient to introduce novelties. I think it's an area worths exploring, but I can't really guarantee it would lead to better challenges.
Keeping in mind that CG business plan is probably slightly more pragmatic than just offering bandwidth and CPU power for free, is it conceivable that CG could make room for some kind of experimental contest? Maybe as a distraction between two real contests, with less requirements (nothing to win and no ranking changes), but also less involvement from CG too (a working example as the last contest referee is all that is needed) in order to play with new ideas?