It's Garys fault!

[nachoman]

This explanation is somewhat at odds with my own understanding. While the axle and wheels are indeed a monolithic component, and rotate as one unit inside journal boxes, the rails are tilted inward by the tie plates, and the wheels are truncated cones that want to ride as if controlled by cam action imparted by their conical section tire surfaces and the canted top surfaces of the rails as forced by the tie plates. The idea is that, as a car rounds a curve, the conical sections adjust to minimized the disparity in rolling rates between the two rails with different radii. The inner tire surface is forced inward, thus making its working surface smaller, while the outer wheel is worked toward the outside of the curve where its tire surface, being now closer to the flange, is greater. In fact, it is on higher speed and tighter curves where the flange on the outer wheel may actually make contact with the flange face on the outer rail, and this is where you would want the rail flange faces to get the grease.

-Crandell

Probably doesn't work on the sharpest curves, and one wheel needs to slip. And, it probably doesn't work so well when the wheels are worn. I have been told that the conical wheel profile is important for keeping the flange from touching the side of the rail. If everything is working properly, the flange should never bump up against the side of the rail. Of course as track and wheels wear - flange rubbing is inevitable.