I agree that eventually a standalone option will be preferable - I just don't think you can make a good enough one now, and I think attempting it is going to mean either huge costs (since you're not reusing current hardware) or huge compromises (ie. terrible performance).
I'm currently running a Vive with a 1080, and I still often can't keep framerate perfect (which you really want) at good supersampling (which makes a huge difference too). Sure you'll get some advantages with a dedicated device and integration, but it won't nearly make up for what you're losing in raw horsepower. Worse, requirements still have a ways to go up before they settle - a HMD really wants ~4K/90FPS at very high quality (well, and very low latency). Even without the extra wrench of 2 perspectives, most normal PC games still can't hit that, even on very large advanced hardware that draws big watts. VR needs more raw power, not less. If there were easy shortcuts to get this stuff, PCs and consoles would be using them; turns out sometimes you just need a bajillion transistors and a big power supply.
While I don't think positioning is the hardest problem for a standalone solution to solve, it is a real problem and you could improve current GearVR solutions 10 fold and still have something that's garbage. Tracking needs to be pretty much perfect or else VR is a barf-fest. Eventually inside-out camera based positioning might be good enough - but it has a tough hill to climb to match current state of the art. SteamVR's spinning lasers and fancy algorithms are magically good - crazy accurate, fast, and even cheap to build. They completely outclass current camera based solutions (like the Oculus uses) even though those solutions are much simpler than inside-out (because it's easier to track a diode you control than random surroundings you don't). Tracking tech really is the magic of Vive, and even a small downgrade could really break the experience.
And yes, currently there's nobody doing wireless video with low enough latency, but that's not because it's an insurmountably hard task - it just needs dedicated work, and VR is a good reason to get around to that work. The hard part - bandwidth - is already there, you just need to trim out some protocol transitions and latency would be very good, perhaps better than wired HDMI by the end (assuming that we can start the wireless chain directly from the GPU). Even current Rube Goldberg setups - HDMI->WiFi->HDMI, often on general purpose computers - are close to good enough. I think this is solvable, but it does remain to be seen whether someone with enough juice (eg. NVidia) will give it a proper go.
Anyway, I'm getting a lot of fun out of VR already, and I think it's going to progress really fast over the next couple years. I'd be pleased as punch of someone came out with a great standalone solution, I just think that's still a ways off.