Mmmm. Gotcha. Never seen those before, very clever!
I'm still convinced that somewhere in Raffs design there is a benefit from what could easily be seen as a flaw...
Those bearing rollers arents actually drivers are they? they are just slave's to an input. They bear a load and they roll with minimum resistance.
Raffs rollers actually exert a driving force when they move position, speeding up or slowing down the surface they are in contact with.
What I was thinking was, if part of the surface of the roller is already at the speed being aimed for, it might act like a kind of viscous clutch for the rest of the roller, easing the speed change. In fact without it, it might require so much force to actually implement the speed change as to make the concept unworkable.
Obviously this is based around the fact that the surfaces of the disc and rollers never actually touch in ideal running conditions (as I believe is the case-Raff?).
They oil is both the lubrication and the traction, so reasonable viscous losses are inevitable with this design.
I'm also guessing that if what I said above is true, trying to change speed too quickly will effectively 'stall' the CVT. If the change in speed is greater than the taper of the rollers and the 'give' of the oil can handle. This situation could be improved by adjusting the taper of the rollers, but would also effect the usable variation in ratio and/or size of the finished product (i.e. longer discs would be needed for the same ratio change if the taper was decreased).
For some reason I have in my head drilling a hole with a 3mm drill bit vs trying to do it with a countersinking bit...
Obviously I am just dreaming all this up, playing devils advocate for what could go wrong, and what could be right about it.
Things like this are why you need a prototype, unless you have some very, very fancy software.