Up front, I'm a Wintel guy who's just scratching the surface of becoming a Lintel guy. I'm no scientist or engineer. I've used a Mac maybe twice in the pre-OS-X days, and never used one long enough to get the hang of it.
But I'm not married to any particular OS or architecture. I'm hearing impressive things about OS-X, and if I had the budget for a Mac I'd want to play with it. I say, if they can build a supercluster based on a G-5, then bravo! The engineering of the G-5 was obviously up to the task. Two things matter in the real world... price and results. Sounds like VT hit their mark, and I'm impressed that they were able to see outside the box and make it work.
Congratulations, Virginia Tech!