I don't doubt that the speed stability of VPI's rim drive improves upon standard belt drive.
Aiming for the correct amount of compliance in a drive system is probably an elusive design goal. With the 300RPM motor on a TNT, I heard significant improvement using a carpet thread-- which has much less compliance than VPI's soft rubber belt. But performance audibly degraded when I tried virtually inelastic Stren braided fish line. The complete lack of elasticity requires that the fish line be tensioned very tightly to grip. It pings like a piano wire and transmits all AC motor vibration to the platter.
From what I can gather, part of the problem is vibration inherent in any AC motor. There needs to be at least some minimal compliance (or "lossiness" as stated by HW) in the drive train to absorb it. In rim drive, noise at the contact point between the driving pulley and the platter introduces additional noise. I believe HW mentioned on AA that the surface contact noise of his rim drive was reduced by using a large-diameter driving wheel. If this is true, then his approach would seem to be preferable to the small-diameter O-ring/motor pulley used by Teres. HW's email suggests that they're looking for softer materials to improve in this area, and also perhaps to improve grip at reduced pressure from the driving wheel.
On the other hand, the quieter DC motor used by Teres would seem to be a better choice for rim-drive. It introduces fewer problems to solve, requiring less compliance(slop) in the drive train than an AC motor. Less compliance = tighter coupling = >speed stability.
It remains TBD which system sounds better. However, VPI's retaining soft belts between the motors and flywheel looks like a kluge approach. While fooling around with thread, I found that even the short rubber belt between VPI motor and flywheel sounded slurry relative to low-compliance thread.
Another way to tackle the AC motor noise issue is through the speed controller that governs the 2-phase AC motor. In this regard I think Mark Kelly's AC-1 controller is something of a break-through: it allows voltage (and other paraments) to be separately varied for each phase of the AC motor. With some tweaking, you can virtually eliminate detectable vibration from the 300RPM motor. In constrast, an SDS(or Walker) cannot do this. With the VPI two-motor cassette, the situation is further complicated by the random phasing activity of two unsynchronized motors.
Of course it's about implementation, not just theory.