Update: I may've FINALLY figured it out.
I pulled the wheels off and noticed that I could shake the brake drums around ever so slightly. That means the hole in the drum must be slightly bigger around than the axle flange. Of course, this is necessary to get the thing on, but since the pulsing is always worse when the drum is warm, I thought maybe there was TOO MUCH play- too loose of tolerences. My hypothesis is that when I apply the brakes, the first pad to contact immediately pushes the drum off to one side, and the the wheel and the binding of the lugnuts keeps it there for the rest of the stop. Being slightly off-center, once per revolution braking force spikes. As it gets warmer, it gets pushed further off-center, and then contracts to nearly-centered again when it cools.
I just happened to have some .002" stainless steel shim-sheeting. I cut off a long strip and shook the drum up and down, then left and right, as I'd insert the shim between the drum and axle. I got it most of the way in- just a tiny bit of overhang. Took a while to work in, and there's still a very tiny amount of play, but a small fractionof what was there before. This means I took up about .004" in diameter off of the hole in the drum, and there's a bit more where that came from.
On the drive in this morning, the brakes felt better than in a VERY long time...and these drums are already slightly warped. I could barely detect the pulsation. Of course, this usually happens for a while after a brake job anyway...but I didn't do a brake job.
If this is it, I may take it back to the shop and ask for some drums that fit properly without being shimmed.
I'll update again if this gets worse, but I think I might have it figured out.
This means I took up about .004" in diameter off of the hole in the drum, and there's a bit more where that came from. Do you think that's enough to make a dramatic pulsation? How far out of round does a drum have to be before it's considered warped?
So, how much play do you generally feel between a drum and an axle?