I have an update. I still haven't figured out the cure... but I think I figured out the cause! This same type of contradictory twist (where the limb leans to the opposite side that the rotation of the tip suggests it ought to) happened to me on 4 bows in a row. I was trying all kinds of things to fix it. It was starting to drive me insane. Now that I've figured it out, it seems obvious. Probably was a stupid mistake, but maybe I can save someone else from it.
It was not that one side was thicker than the other. It was caused by the way I was wrapping bike inner tubes around it during the glue up. I had been wrapping it directly around the bow, like this:
When I started wrapping around a form as well, like this:
The problem went away completely.
I don't think this is due to the support board keeping things in alignment while the glue cures, as I was always careful to do that and my bows were well aligned and supported. I think it has more to do with the wrapping process itself. While wrapping around the bow directly, the bow is getting torqued with each wrap. The more muscle power I put in to pulling on the tube to get it tight, the more the bow was getting torqued. When wrapping around a support board, the bow is not getting twisted like that, as the inner tube just gets laid over the bow like a seat belt on each wrap.
I'm still a little puzzled, as it's interesting that torquing the bow while wrapping it caused no visible twist in the unstrung bow. Yet quite a lot of twist once the bow was bending. It must have just caused it to glue up with some weird internal stresses. I'm still trying to work out the physics of that. But the support board definitely seems to have fixed the problem that was plaguing me. There's a moral in there somewhere.