FWIW, I have been doing little in the way of electronic projects for the past year or so, and I stopped working on the reverse engineering of these devices about 2 years ago.
I now have a new breakout PCB for the 44 pin EPM7032 device. I am currently trying to re-collect all the various places on my PC where I did work on this project and I hope to spend some time trying again, with the hope that the new PCB will be much more reliable than my old "mess of wires". I now have the breakout PCB wired to the original MEGA 2560 Arduino, with pins in the sketch re-mapped accordingly. After some wierd behaviour of the two DACs, I discovered that my original "play" devices were both broken/shorted causing the DACs to mis-behave(!). Using a new device, I can once again read the "device ID" and get the magic "ALTERA92" string which confirms that the device is correctly in the test/programming mode (Vpp > 14V).
Incidently this data is stored in one of the many (60+) blocks of EEPROM. I am pretty sure that when the device is erased this block also gets erased as well. From recollection during the erase the old "device ID" is read, before the device is erased, and then the "device ID" is written back to the same specific EEPROM block. This is necessary (at least for the ALL03) because for any operation, the ALL03 software always tries to read the "device ID" and if it fails, it refuses to do anything else. Two bytes read from the special "device ID" block contain information about Vpp and the pulse width needed to erase/program the device. The software has a useful table, used to decode and print the information, held in ASCII format (C null terminated strings).
I am trying to re-visit the work I did on the AMAX70.EXE software using Ghidra and IDA-Pro. Both were very helpful tools, however both were far from perfect, and have a steep learning curve to get the most from them. Out of curiosity I installed the latest version of Ghidra, hoping to see improvements, but the code re-generated was worse in the newist version as compared to the old. This might be fixable, but the menu options in Ghidra seem less numerous and flexible than in IDA-Pro.