We build out of a snowmobile repair shop so we have a great deal of hands-on knowledge with this sort of technology hanging around the shop (pretty much all snowmobiles use CVTs of this type).
First off, the traditional design does not function in reverse as you depend to a certain degree on the torque of the belt to influence the movement of the secondary sheaves. It would require extensive reworks of the helix and ramp angles to get it to even function marginally in reverse. (I have a friend who built a gokart and tried mounting the secondary running backwards due to space constraints and we could never get it to shift properly).
Secondly, the losses with a belt based CVT are fairly significant for an FRC application I doubt you would want to incur them.
The comet system is designed for engines from 5 to 8hp. To use it on a low power application would require massive reworks of the springs and weights inside. Again we come back to the power losses once you did get it to work. Those comet belts are fairly beefy when you're talking about CIM motors. Also the ratio I believe varies only from 1.6:1 through to 0.89:1 which isn't too much of a spread.
Typically, the purpose of these types of CVTs are to keep an engine running within a very specific power band. Gasoline engines, especially the 2 stroke variety have a very narrow range of RPMs where they produce peak power. CVTs of this type are designed so that the weights and springs which control them are at equilibrium with each other when the engine is at a specific RPM. Any faster, they upshift- any slower, they downshift to balance the forces within the sheaves. The result is that the machine is always running as fast (Given WOT) as possible by keeping the engine in its optimal power band. With electric motors, you don't exactly have the same goal since their powerbands are vastly different. You can trade amperage and voltage for torque whereas on a gasoline engine you do not have the same luxuries. It would take extensive testing to develop a purely mechanical system which optimizes the power characteristics of an FRC drivetrain. For this reason I would say an active system controlled by encoders, current sensors and programmed logic would be vastly superior and much easier to tune.
If you were to do away with the mechanical weights and springs in favor of a servo motor to control the sheave separation you would solve the reversing issue. If you were to downsize the whole thing and use something like a silicone O-ring as a belt running at high speed to reduce your losses, I think you would have a viable system.
All this being said though, I think you would enjoy much improved reliability, reduced complexity, weight and improved efficiency to just pick up some ball shifters. I don't know that the benefits would outweigh the drawbacks. The again I have always thought it would be cool though to run all your motors at full speed and control the drivetrain through an infinitely variable transmission like a hydrostatic system
