I don't understand how this thing works either. If you have a torroid partially fluid filled, and rotate it eccentrically, the fluid will move to the area furthest from the axis. It has to, the same governor weights move outwards against the spring as they are spun, and fluid flows downhill.
Sounds like Snake oil to me, but if people say they have measured an improvement to validate it, obviously I'm missing something. Perhaps there is a dampening effect due to the viscosity of the mercury in the tube? If that's the case, when the unbalance is rectified by the movement of the mercury to the light side, what keeps it there now that the assembly is spinning concentrically?