I'm not at Railsconf this year, but before he left I asked my coworker Krishna to beg for, borrow, or steal a copy of the Maglev alpha code and bring it back to the office. Unfortunately, it looks like that's not going to happen (in an email he says "They don't have anything to share at the moment"), damn it.
Still, as Obie says, this is a huge deal. Even without the stated performance improvements that they're already seeing (I'm hearing 5x to 100x in microbenchmarks vs the MRI), it would still be a huge deal, because this might be our best chance ever to move people to a real object database. And not just any object database, but one of the best object databases out there.
Once you use it, you won't ever want to go back.
As far as I'm concerned though, the bigger news about this is that the Maglev VM "does retain the ability to run Smalltalk code." As long time readers of this blog know, I'm a big fan of Smalltalk and the Seaside framework, and being able to start a project in Rails and then use the same data to build a Seaside app will knock down another huge barrier as far as getting people to try Seaside goes, (see my my post from a year and a half ago about sneaking Smalltalk in through the Ruby backdoor).
In my experience, Seaside makes writing web applications faster and easier than even Rails does.
Paul Graham is well known for saying that their secret weapon while developing Viaweb was Lisp, because it let them develop software faster than any of their competitors.
Ruby + Smalltalk + Gemstone is your secret nuclear weapon.
Not only will it let you develop software faster than any of your competitors, it will also let you scale that software in an easier manner than ever before.