Yeah that's not a very good animation, I've seen the same elsewhere on the web, I think it's a copy of a copy of a copy.

I haven't ruled out that you're right, by the way, but that just seems contrary to rolling bearing theory. But I need to prove it if I'm right or wrong.

I think that you believe, that the rolling contact speed between the spinning balls and inner race, must then be the rolling contact speed between the balls and outer race, because it's the same ball, and the outer race is fixed, except that can't be because the outer race is larger diameter. But in addition to spinning, that ball is also translating, and that, I think, compensates for the contact speed differential. I haven't quite got it mathed out yet. What I got is (for this specific animation):

For each 1 revolution of the inner race, each ball rotates 6 times. That means a 6:1 ratio in diameters. To get the outer race rolling diameter, it's 6/6 + 2/6 = 8/6 or 4/3 times the inner race diameter.

That's it so far. I'm straining my brain to figure out how to take into account the translation of each ball, and whether added or subtracted to the rolling contact due to the balls spinning. If it were a linear bearing, rail moving balls over a fixed rail, it's easy, it's the "2 for 1" moving rail translation versus the ball translation. But with diameters, it's going to take more thought.

EDIT: I think that 6:1 ratio above based on rotations is way off. The race diameter ratio is a lot closer to 3:2, with balls at 0.5 diameter.