My guess is big players--Battlestar Galactica Online, Tiger Woods Online--will use the Flash player export as a fallback. Depending on its shipping performance, and how demanding the content is, developers may even fork the content for each player individually. Regardless of multiple versions of content, it would probably make sense to push users into downloading the Unity Web Player for better performance.
Unity is also working on a NaCl port. In a year, 3D web game developers may be able to offer the Unity/Flash/NaCl versions simultaneously (prioritized by browser/performance/compatibility).
In general, Unity is well-positioned to exploit other platforms as they emerge. If they can manage this Flash export, they can probably manage anything...
[Full disclosure: I'm a long-time Unity user, and I help them organize proposals/sessions for their Unite conference, but I don't actually work at Unity]
I'm not sure if that's possible on Flash or not, but if so it'd be easier to bring your performance up to the platform limitations...
I assume the main reason is that 3D in Flash (Molehill) is very recent? Now that Flash has that capability, you can write a game that looks like that in Flash directly, Unity is not strictly necessary (although it might be a useful game engine on top of Flash).
On a side note, it would be nice if Unity produced iOS apps that could be extended as easily using Obj-C / CocoaTouch the way it looks like you can work with native AS3 code.
I would agree, this is a good looking game.
I honestly don't know much about Unity or the technology but my first concern is performance. Adobe tried the "write once, run everywhere" and while they were successful, performance suffered. How will something much more complex and demanding perform?