ClearType didn't come out of MSR, Kinect was an acquisition, SongSmith has gone nowhere since it's mild viral success a couple years back, and Photosynth was based on an acquired project and company.

F# might be the only item you listed that originated in the MSR division and has moderate success.

This video http://techtalks.tv/talks/54443/ on machine learning by Bishop talks about the research they applied to kinect. There was much work done beyond the acquisition.

Anyways watch the video, its real good and worth watching for its own sake.

Of the things you mention, the only successful one is the Kinect, and it doesn't come from MSR: it was an Israeli company that Microsoft bought after they were turned down by Apple.

Speaking of which, do you remember that about the same time that the Ipad came out (or just before then), Microsoft was demonstrating a touch surface for coffee tables?

Microsoft Research was never a product development arm, it was a research arm. In that area they've succeeded magnificently -- a huge amount of seminal OS papers (back when I was doing systems research) came out of MSR. In contrast, the Google papers always were (and I think mostly still are) complete crap. I say this as an exiting Google intern who's trying to work for them full-time.