Originally Posted by
ctmullins
But I cannot, for the life of me, figure out how a sprung upper pivot improves things.
It really doesn't. What it does do is increase the total chain take-up capability by letting the entire cage move forward/backwards as necessary. Of course the actual behavior depends on the balance between the two springs, and as far as I can see that is pretty unconstrained. The moment-to-moment orientations of both the cage and the derailleur body also depend on the friction in the pivots. So in general it sort of works, which is sort of okay, just not as well as the Suntour design which constrains the motion precisely. On the Suntour design one could wish for different slant angles to match different cog sets, but even with their predetermined angle it worked better than everything else.
The Crane/Rally technique of orienting the two pulleys and the cage pivot in an "L" shape makes the jockey pulley move downward/upward in response to shifting to a larger/smaller cog. This is good. But the pulley also moves in response to shifting between chainrings. This is not so good. If the chainrings are close together it isn't so bad, but closely-spaced chainrings isn't in keeping with the desire for low gears. From experience I can say that even with a half-step crossover pattern (47/42) and a widely spaced 6-cog freewheel (14-34) a front shift can sometimes change the angle of the chain at the rear enough to cause a spontaneous rear shift if the RD isn't perfectly centered.