@gaucho777, thanks for the T-mar quote.

Some of those points make sense but may not be so significant.

I have only one bike for which the handling is, ah, quick enough that I'd care to lengthen the rear. (And my ego won't let me anyway.) After a weekend on the tandem every solo bike I own feels like a twitchy sonova' for the first mile. After a few days commuting on the Grandis even its handling begins to feel normal.

For chain alignment, I don't see that it matters. A 1cm change in the screw position will give you a variation in the effective distance from the top of the selected ring to the top of the selected sprocket between, say, 46 to 47cm, or a bit under 2.2%. If the FD moves, say, 6mm (the nominal sprocket separation applied to the rings), that 2.2% is equivalent to re-aligning your BB by only 0.13mm, essentially insignificant. Even if I'm off by a cumulative factor of 3 this is still less than 1mm re-alignment of the BB.

I'd argue that any generalities about the guide pulley/FW position apply only to specific cases, if you take my meaning. On different RD's the guide pulley moves differently as chain take-up happens. On some the pulley moves fore and aft, on others it moves in a tight circle, on others it is mostly vertically.

With a Rally/GS style cage, the movement is almost entirely vertical by design, giving a similar effect as a slant parallelogram. But unlike a true slant-p design, the vertical position of the guide pulley changes with a front shift as well as with a rear shift. To put it another way, a pulley to sprocket distance optimized for one ring will not be ideal for the other. Since it must accommodate the higher position when on the small ring, the big ring will always have a greater distance than optimal. The bigger the front shift the worse this problem will be.

Which brings me back to the subject of this thread. When I first set up the Masi (Henry III long cage on a NR body) it didn't shift well. I concluded that the pulley-sprocket distance was too great. A larger guide pulley helped some but of course it occupied slightly more chain, moving the pulley down and negating some of the gain. Adding or subtracting chain links meant changing the length by an integer number of inches, not what you'd call subtle.

By far the most significant effect on the drive train from dropout screw position is effective chain length. Move the screws, say 1cm, and you change the effective chain length by 2cm. That's 0.78". With a Rally-type cage that has an arm ratio of about 3:1 that raises or lowers the pulley by a bit more than a quarter of an inch. That let me fine-tune the vertical position of the guide pulley to a good working distance for the big ring while still not too close for the small ring. The fact that the rings are 47-42 helped, of course. That the ultimate axle position came out to be forward anyway was an added benefit to my ego and street cred.

The Grandis has a similar setup (Soma long cage on a NR body) but I'm not sure about the cage's arm ratio. Its gearing is different and maybe I lucked out find in a good screw position, but it shifts just fine.

Anyway, for this type of RD cage the vertical position of the guide pulley is the real test, and the only reason I've found for tweaking the screws one way or another.