I think I may see how this works. Unfortunately I'm going to be swamped this week, I think, and I won't have time to write it up sensibly with pictures 'n' such any time soon, but here's a quick sketch.
The key is the outer race. We've got 100 amps flowing through that race. IF the current were going unidirectionally, it would produce a B field inside the bearing which would, given the radial current through the balls, make the bearing spin. (In fact this would be exactly like a unipolar motor.) But the current is going to split, going in one direction on one side of the race, and the other direction on the other side, and the overall effects cancel; the average current direction in the outer ring is null, which is why it doesn't self-start. But that "null average" ignores the INDUCTION of the race, and the fact that the balls are moving and the configuration is changing. That's the solution; the rest is details. Herewith some details. Now, assume there's just one ball, for simplicity, and just one point of contact with the race. Assume the contact point is positive, and is at 12 o'clock, and the ball is at 3 o' clock. Then there will be something like 3x as much current going down the ring to the right (from 12 to 3) as there is going down the ring to the left (from 12 to 3 'the long way'), due to the resistance of the race material. Keep the contact point at 12 o'clock. Move the ball down to 6 o'clock. Now the current splits evenly. Move the ball around to 9 o'clock. Now 3x as much current is going down the left side as the right side. If the bearing is spinning clockwise, then the current going to the right must be DECREASING as the ball goes down around the bottom of the ring, and the current to the left must be INCREASING. But the ring has inductance, and with 100 amps total going through it we can't neglect that. The inductance will tend to resist current change, which means the current pattern is going to "lag" what a static analysis would lead us to expect. In short, with the bearing spinning clockwise, and a positive contact on the outer race, there will be more current going CLOCKWISE than we expect. In other words, the average current direction in the outer ring won't be zero, it will be clockwise, and as a result there will be a net B field in the ring, pointing INTO the ring as we look at it. With this configuration, with a positive contact on the outer ring, the current in the balls is going from the outside to the inside, and the force on the balls due to the (net, average) B field from the ring will also drive them CLOCKWISE. And so the ring will continue to spin. QED (I think!). Could this be tested, using multiple contact points to the bearing, maybe? Not sure. OH, yeah -- and what about the need for magnetic material? Well, if I'm not mistaken, a ferromagnetic race (and adjacent balls) will increase the inductance of the race versus nonmagnetic material. That in turn should make the motor work a lot better, if this explanation is correct. (And didn't somebody cite a source indicating that the motor might work, *some*, with non-magnetic materials?) I hope this is clear enough that folks can follow what I'm trying to say. In any case I have to go back to bed at this point. 'Till later....
