I dug into this last night a bit with the scope. Here’s what I found:
When I disconnect USB from the computer, I have no regulated power to the LCD, and I assume, the uC as well. I have what appears to be raw power from the transformer with an RC component from the filter caps. Even over the 5v rail. So it looks like there is something up there. So, for the rest of the discussion I will always have the PC attached.
When I probe the signals to the LCD, there is an incredible amount of switching noise, I expect from the chopper drives. That explains why the LCD changes behavior when I touch it - my noise contribution plus all the other noise allows the LCD to read the data differently. High impedance inputs just love to couple to noise sources.
Oddly enough I do seem to get a clean reset every now and then, but never regularly. I am using various 20 x 4 LCD panels in my tests. I have used this kind of panel for over 15 years now, so I am very familiar with them. Generally these are a cakewalk to interface, because they are so undemanding with regard to timing, but the amount of noise I see would seem to explain a lot.
I am starting the process of shielding the case and adding ferrite to the cable to try and properly manage the radiated and conducted noise. The tests I did last night started to confirm that the pin settings are right, but as far as signal - well, it’s too messy to tell…
Thanks for your assistance, and I will post any changes.