Having loaded A_05 into both failed units, and running system self-tests, I see the following failure:
RTC RAM TST Address Bus Failure.
I'm assuming RTC - Real Time Clock, and that this failure references the Dallas chip. If this is the case, I'll put down the project until the replacement chips arrive and solder those in.
I've seen enough of these dead chips to believe a dead battery internal to the Dallas chip won't cause this, as all the units I've seen with dead batteries function just fine, but don't retain programming information on power cycle.
I think I solved my own problem:
Yes, the chips can throw a "RTC RAM TST Address Bus Failure" if the battery's completely dead, missing, or in my case, the replacement battery isn't actually soldered to the internal pins exposed by dremeling.
The lead wire I used must have just melted to the plastic housing mimicking a solder joint. That was the first unit. The second, I'm sure I tried to test it without soldering in a battery first.
Both units now boot, have successfully updated to the final version of firmware, and seem to be chugging away.
FWIW, I was able to grind down the chip to expose the battery leads with the chips in situ, even leaving the motherboard attached to the bottom of the chassis. Much easier than removing 20 pins' worth of solder from the PCB. I'm using a CR2032 battery/holder hot glued to the top of the chip, and attached to the solder points with small jumper wires. "Pin" 16 on the chip is - and "Pin" 20 is +. I say "Pin" as this pin is bent up from the internal chip and connects to the internal battery; these do not exit the chip housing from the bottom.