spyrestudios.com
The difference is between something designed to impress (like, say, a ceremonial sword) and something that actually does its job efficiently, every day (like a kitchen knife). The biggest, most popular, most used sites in the world all have really simple, get-out-of-the-way designs: Google's UI is almost not there, Facebook's gets simpler with every redesign, Flickr's tries hard to disappear completely.
Now, there are some objects, very rare, that are both beautiful and used every day -- the kitchen knife that's also pretty. I'd put Flickr's UI in this category. But the majority of heavy-traffic designs are ugly but functional: think Craigslist, Amazon, and our own Hacker News.
So no disrespect to these designs: they're amazing. But sheer visual impact is not the only metric for judging web design.
They look like print magazine spreads.
Which goes hand-in-hand with your observations on their bright-but-short impact.
On more than one occasion, I found something I liked there and I had a paying client who wanted "something like it"; either the artists weren't answering emails, or they were busy doing term papers, or they just sent me nice step by step instructions on how to do the same in photoshop along with the plugins they used, yet some "added" me as a friend and forgot to reply.
I agree though; there's a ton of talent on that site, but most of those folks aren't really out for money.
A) An edge site(being personal, small community, niche audience)
B) A commercial site trying to get the "hip young crowd".
One of my top concerns when developing sites is the size of graphics. The bigger concern is the size and complication of javascript. Implementing sites like these would be fun, but what kind of overall download are we looking at for our users? How much does this affect our sales or ad click-thrus for our audience?