Lack of a sync API I'm guessing is a big one. There's no way to guarantee a file you place in a Dropbox will be there in the future unless you're notified if it's gone (which you can now do with /delta). Also, storing files isn't that big of a challenge, it's not like Dropbox is equipping you with a new skill. There needs to be a very compelling reason to put someone through an OAuth flow that requires a third-party service subscription for something that can easily be done transparently on the server or using S3 directly.

I personally don't think it's compelling for everyday file storage, like S3 is. However, I think there are a ton of other possibilities that are very compelling: data export, save to Dropbox, collaboration and sharing amongst multiple users, etc.

I'm a lead on The OpenPhoto Project and it uses Dropbox as an optional datastore. They're not really a CDN though so it's paired with S3 or the local filesystem for CDN purposes. http://theopenphotoproject.org