I expect a risk that people will vote on something very nebulous (more meta! more meta!) as the best idea, and at the end pretty much nothing measurable happens.

Maybe there's a risk, but do you think it's big enough to make the Hackathon not worth doing? I think that people here are smart enough to Get Something Done.

To avoid going too much meta, I would recommend adding an artificial constraint, such as "we must be able to complete the whole thing in one week".

Hmm, maybe.

1) I think that if a private group wants to do something more future-oriented, they should be allowed. Maybe you could just restrict people that enter voting to be things that could be done in a certain time period.

2) "You have to have a useful version in x weeks" is probably better than "You have to have completed your project in x weeks".