If you're an interface designer (I am), you should pre-order this thing. This will be a standard form of interaction in a couple of years and you should jump on it early and start figuring out the kinks. Too cool.
There are so many use cases where it doesn't even work that would require a complete rethink of how anything is presented on in internet. For example how about the HN comments where it's pasted as code and it scrolls horizontally. Target that and scroll it with that system. As soon as you have a single use case where a mouse and keyboard is more effective you blow your value.
Moreover, it's not enough to just copy our current understanding of UI over to a form of input like this. Of course that's the natural inclination, but in reality, interfaces will change and adapt to things like this (meaning scrolling may not exist and a whole new form of pagination may be invented). The "how" is up to the designers.
More than anything, you really articulated my point by saying that "it really won't." You're right: as things stand in terms of interaction, this would become tiring. You just have to think of a way to make it not.
They do claim sub-mm accuracy; maybe applications in the small are realistic.
So instead of arm-waving, think of rotating your hand just above the touchpad to rotate and object in 3d space, but briefly. And the touchpad would still work like a regular touchpad, but maybe you don't even need to touch it.
Sub-mm accuracy seems to imply that really subtle gestures could work.
Disagree about the physiological concern. Provided your elbows are resting on the table, it would be easy to get used to. Humans would adapt and it would be healthier than our current much talked about static postures.