I think Shuttleworth's basic premise is completely sound -- I know from experience that watching actual users, in the field, using my 'well-thought-out' and 'intuitive' software, found it much less intuitive and well-thought-out than I did. It was a valuable (and, yes, painful) learning experience. The problem I see is that he suggests putting the software in the hands of user interface experts, which are even worse than developers at predicting how real users work.
I've read lots of articles written by usability experts that talk about idioms, efficiency, doing what the user expects, etc. etc. And that's all great, except that it's wrong. If you follow the experts, then you get in a habit of saying "users are used to software X, so if we want to make our software easy to pick up, we'll make it use the same idioms and concepts as software X." Out in the real world (yes, I worked there for a while, and yes, it was painful), people are asking themselves, "why, for the love of god, does every piece of software do this same stupid thing?"
In every conversation I've had with a well-educated, progressive developer about usability, the argument they believe is that users expect the status quo. (I've been shot down from making a few small changes to our software's user interface on these grounds.) When I've asked everyday, non-technical people who actually use our software what they would think of the change, despite the fact that it behaves differently than all the other software they use, over 90% of them have said "oh yes! That would be a GREAT change! I'm so sick of programs always doing it the other way!"
Now, I'll be the first to admit that what users say they would enjoy and what they'll actually enjoy are often different. But I'm inclined to believe them, because I too have been in situations where I constantly have to fight with an interface choice even as I watch other software rush to copy it. I know that experience first-hand. And I know how nice it would be to find software that stopped listening to the so-called 'experts' and actually spent some damn time thinking about what the best design choices would be.
The Linux/OSS community, as a whole, has spent a long time and a great amount of effort trying to become as Windows-like or Mac-like as possible. And I'm not going to say they/we shouldn't have done that, because one thing that such similarity does is it lowers the learning curve for people who want to try Linux/OSS or switch completely. Both Apple and Microsoft have also figured out how to focus their development efforts on the user, instead of developing for their own use, and that's another lesson it's good for the Linux world to learn. But we shouldn't be afraid to ignore the way the commercial OS writers are trampling and take some time to consider what the BEST direction is. Too much regurgitation of the accepted "expert" design, and you get into a feedback loop where you can't make a better design decision because you've written too much stuff that behaves the "expert" way. And then you're Redmond.
So should we put our software in front of the users? Absolutely! There is no better, more accurate, more incisive, more disillusioning gauntlet for our software to go through. But screw the "experts." Seriously.