Ok, just a little background: I am a senior physics major at Hillsdale College, and have been working on this exact problem for my senior thesis. Results have been mildly conclusive, however, a few things worth mentioning:
The paper which this article refers to contains very few details which are sketchy at best. It contains precious little data about the magnitude of the rotation observed or the factors which determine it. While this does not prove them wrong by any means, this should invoke extreme skepticism already. Hence one of the reasons I have chosen this for my thesis.
Now, I have not completed analysis of my data yet. However, I have come to one very important conclusion: The orientation of a sphere effects the amount of rotation. I do NOT mean that the orientation of a sphere -with regard to the other two- influences the rotation. If you remove one of the spheres and then proceed to reattatch it to the same position in space but rotated from its initial orientation (Say, 1/4 a turn or so), the amount of rotation of that sphere is effected.
What does this mean? If I had more time to work on this, I'd love to dig deeper into it, but the most obvious conclusion is that the rotation is not fundamental to a perfect sphere, but rather is due to small imperfections on real spheres. You cannot make a sphere which is perfectly smooth. There is always some error, however small. With large voltages being applied (in my case, 3-12 kV... in their's, I believe it went as high as 20 kV, although I am not positive), even small surface imperfections could cause small forces. Given that the magnitude of rotation is very small (About 1/16 of a full rotation, at most. On the order of 1/400 of a rotation at least. Given the hair-thin Tungsten wire used, this is a tiny force), this seems quite plausible.
However... I should really be working on writing my thesis instead of talking about it... so... away I go!