Hmmm. I modified my keypad test program to have a variable delay and actually my 7800 appears to be on spec. The keypad requires a 400 microsecond delay and it starts to function correctly between 120-150 cycles (342 - 428 microseconds). Of course I only experienced the voice problem in dungeon stalker when I triggered at least 3 phrases at the same time so maybe being on spec is why.
I've attached the test for people to try. Keypad in second port naturally, joypad in first port to change the delay. I failed to correct a bug where the delay shown is 1 higher than what's used due to the order of checks but hey, it gets the point across.
If my hypothesis is correct then the more susceptible to the 7800 is to the voice issue the higher a delay it will require to read the keypad correctly.
The right side of the pad will tend to work fine for the most part, as you increase the delay you'll start to get more accurate readings, but as you press multiple keys you'll get brief false readings flash up. When you get the correct display all the time is when you'll want to make a note of your readings and compare it against another machine.
Just had the issue again. It doesn't seem to be as severe on my machine in that the avox doesn't reset. I fought the Wizard once with no problems, then the next time after moving around the screen a few times it said something like "I have advanced" followed by a brief pause, then "aaaaaaaaaaaaaaaaaaaaaaaaah", then he took off in his roflcopter which goes "tuh tuh tuh tuh tuh tuh tuh".
Posting my question on this in the other thread here: