Hehe, actually one can produce a very high speed out of the steppers, based on some factors about the load, motor, microstepping, and how your velocity curve is implemented. I've been able to push that setup as fast as 5-7KHz step rate - but at that point, the metal worm gearing starts to howl and scream, and the amount of grease required to support those speeds results in too much backlash. I did a test last night with some new asynchronous motor code I'm working on, and pushed it to 10KHz, but that was stressing the gear chain too much.
The reason why the tilt is not centered with the mass of the camera is primarily to reduce the size/weight of the unit, while still allowing maximal tilt. In this case, with the camera off, the tilt bar can rotate a full 360'. Of course, one can simple turn the tilt upright frame around 180' (it's just two bolts that attaches it to the pan base) and get 360', but one goes completely off-center of the pan.
For those shooting panos, they'll want to be able to nodally-center the camera, so in that case, you just get a longer piece of extrusion to mount the camera to, and then slide it down a bit =) For timelapse, a little off-center is better, as you mention, to give a feel more like a person panning the camera.
FWIW, those motors only have 9 oz/in of torque! The gear chain is 120:1 worm gearing, and power is cut to both axes between moves in that video. So, it can run on a 4Ah battery all day and night - and then some.
!c