My working assumption has always been that frame alignment was
not a factor. I can't explain how I came to that conclusion exactly, other than knowing some frames that shimmied were near perfectly aligned.
I'll leave "perfectly aligned" undefined here since FBs don't totally agree on what that means, but the bikes I'm talking about were fastidiously aligned on a top-quality pro framebuilder alignment fixture (Marchetti) by a competent experienced operator (me). So at least by
some definition, the frame was near perfect.
Paul Brodie, in his video on frame alignment, complains about the fixtures like the Marchetti that use the BB faces as the datum. He says something like "nobody ever flips them over to see if it measures the same on the other side". Well I'm here to tell ya, some people do flip them over. Repeatedly, habitually, a lot. I'm intimately aware that the BB faces aren't even parallel to each other, (yes, even after facing with a Campy tool), but that doesn't mean the BB face can't be a useful datum, it just means you have to be aware of its limitations. But I digress. I assert that different ways of measuring a frame all result in the same thing (a straight-riding bike)
if they are all done correctly — I hope that's non-controversial.
One frame that comes to mind, we sponsored a top Swedish amateur team, just to the extent that we sold them a bunch of frames below wholesale, not a gift. They were in the US and got to ride the bikes while still here. Their biggest guy rode a 66 cm, which was not a custom, just the largest size our stock lightweight road frame came in. We had made literally thousands of those frames, nearly identical, but this one shimmied. We tried many things, like heavier wheels, forks with more and less rake. Nothing worked. We triple-checked the alignment, it was spot on (by our definition).
That was back in the era of 1" toptubes on everything. Larger frames got thicker wall tubes but the diameters stayed the same, due mostly to lug availability. We had just gotten our OS lugs and tubeset (designed by us in collaboration with Takahashi and Tange), which had only a 30 mm DT (slightly OS) and 1-1/8" TT, so I quickly threw together a "team issue" custom for him. (For us that meant as close to stock as possible, and with a lower level of aesthetic polish, i.e. left a little rough here and there). Angles, lengths, wheels and other parts, and his position on the bike were identical to the previous bike, and the frame weight was about identical too, but the new one never shimmied.
All frames from 62 cm and up got OS TTs from then on. The DT stayed at 1-1/8" though, for that light road race model, even on the 66 cm. (The full OS tubeset and lugs was for a more crit-oriented model.)
I know for sure I can't analyze shimmy in my head, but I
think I have noticed a positive correlation between frame flex and shimmy. Maybe with low trail being a secondary factor. But the one thing I am pretty sure about is that it's the
bike + rider system that shimmies. The same bike that shimmies with one rider aboard often will not when someone else rides it. For example I wasn't able to get that Swedish racer's bike to shimmy for me.
Anyway I'm not nearly sure enough about any of it to rule out
unterhausen 's theory that it's caused by poor alignment, as long as he'll allow for exceptions, frames that still shimmy even though properly aligned. Which would mean there has to be more than one cause.