There is an interesting part of this video that illustrates just how hard multi-user multi-touch is to do right: Note that the scene of the girls "fingerpainting" is faked.

Why fake it? Drawing programs are very modal: You have one brush, with one set of properties. You have one selection you can move or operate on. One property sheet to edit. Etc. Except for games with their own interaction framework, most OSs assume one user with one selection in one document at a time. It may seem like a trivial change, but true realtime collaboration on a single screen seems to demand a UX system designed to support that.