Fixed!!
R-53. A 22K 1 ohm resistor that develops the bias for the noise inverter stage was open. I didn't (still don't) understand quite how the noise inverter works, but once I remembered how to use the Line trigger mode on my scope to get decent traces, I walked through the sync circuit and confirmed this is where I was losing the sync pulses. I found that the cathode voltage was nowhere near where it was when I first checked it a few days ago - but that was when the problem was intermittent, so the resistor must have been changing value at first, and then finally failed completely. This explains why it initially would work for 15 minutes or so, and as the resistor value went up it finally reached the point where the sync was lost. Sync is solid as a rock after an hour, so I'm declaring victory!
Thanks for the helpful suggestions from everyone here. I definitely understand B&W TV circuitry better now, which is a good thing as I think I've got another set coming my way.
I'd also mention that I found a copy of "Television Servicing" by Alex Levy and Murray Frankel on my bookshelf, and it seems to be a very good resource. While the Predicta circuitry seems to be a bit unique, I found a lot of useful information and tips in the book