Flat surface area exposed to the wind is going to push the bike around than turbulence around a protruding bit.
For me, I think it is about energy transfer.
It takes more energy to absorb or convert wind energy to turbulence than to simply deflect it with less turbulence.
There are lots of "protruding bits" in an unfaired bike. Which is also a reason why it takes to much extra HP to push the bike to higher speeds on an unfaired bike. Side loads are no different, aerodynamically, the more bits that can be stopped or catch the wind and convert it to turbulence the more kinetic energy is taken from the bike to change the wind behavior. This manifests itself in un-commanded side motion of the machine.
Shape matters. The fully faired shape must present a neutral center of force to the machine and the oncoming airstream in order to keep the steering neutral.
I think it also matters how the windshield fairing is attached. If the fairing is attached to the frame, like the Goldwing, it is much less likely to cause an issue.
Add-ons that attach to the steering column do cause problems as the side buffeting becomes steering input.
Quite true. IF the wind load transfers directly to the steering mechanism, both side load energy and whatever is transferred to the steering mechanism effect stability.
When I put the Vetter windjammer on my CB550, I dreaded all the horror warnings about crosswinds. When I eventually encountered high cross winds, it was certainly no worse than before. And, if anything, it felt even more stable. The Vetter IS frame mounted. But, it isn't a full fairing, though. And at the time I still didn't have the lowers on it. But, if it was weather vaning by absorbing more energy at the rear of the bike than being deflected around the fairing, it actually helped keep the bike stay in the lane rather than simply blown into the next one.