Which is not to say that your concerns don't have merit, but do note that the success of Diaspora and the ipod wristwatch projects seem very much to be outliers. Kickstarter claims that a little under half of all total projects /approved/ actually meet their funding goals, and from a glance of the site, it seems as if most successes are rather /modest/ successes - we really only notice the outliers.
The site also does warn project organizers that failing to deliver could potentially leave them legally vulnerable. Given the relatively low success rate, the human curating and the necessity of maintaining a public and accessible persona to drive up donations, I wouldn't call it a great platform to scam with.
Of course, given my limited experience with scams, I could be totally naive about this.
You should have to set a maximum as well as a minimum on a kickstarter project. The first time one of these massively over subscribed offers turns out to be a total scam, or just fails there's going to be a lot of unhappy people.
I think pre-sales projects like this are really interesting, but it doesn't really match the donation model. At a minimum, kickstarter should escrow the money and release it in $xx,000 chunks or something, and ensure people are getting what they ordered. Even better would be to not have people paying until they were reasonably close to an actual delivery date. It's one thing if you're waiting for upfront R&D, but it's another thing if you're simply on the ass end of a one year waiting list.