I won't notice better than 48 hz, except how it impacts my head tracking. 60 fps versus 75 fps when head tracking is involved at least appears very different. There could be an issue with some factors not being done perfectly right or an unfortunate interaction with the motion sampling frequency that could change, but at least for now it looks night and day to me.
I/O on a server that can process that many inbound channels is a biggie to overcome without a massive back plane.
I guess I'm a bit perplexed at this statement. The current DK2 uses two USB 2 ports (for camera data and motion tracking). A tiny fraction of modern IO capabilities, and it seems to do just fine. Sure it only tracks the head, but technology adding more tracking seems content with USB as a bus. I'm not sure what IO load you are referring to that would be infeasible with modern systems.
One object does not give anything meaningful in VR,
Well, just head tracking is *huge*, though the lack of positional tracking without augmentation is an issue. My point is those sorts of sensors, which are a valuable component of comprehensive motion tracking, are now commodity items. DK2 provided good experience using camera based tracking to what amounts to a somewhat higher end version of common motion tracking equipment.
In general, you make it sound like you haven't been involved in the market since RH ES3 days. I had the opportunity to experience VR demos back then in a few instances, and it wasn't compelling. It's night and day compared to the DK2. I haven't had the chance to demo modern high-end VR simulators, I can't imagine how those have changed since back then.