A1SCPU with A1S64AD- it works, it doesn't, now it works again

[gurar]

I work for an OEM that has used the A1S in one of our products for 15 years. I'm very happy with this PLC and sad to see it go. Occasionally, a customer will neglect to change the backup battery and must send the CPU in for a program reload. Recently, one came in and my coworker took care of it. The battery had been changed a few years ago, and this seemed unusually early for a battery-related fault. After the customer reinstalled the reprogrammed CPU, his machine had a problem that led me to believe the A/D module was not working: good voltage inputs, but no response in the machine parameters that use data coming from the A/D module. Otherwise, the program and CPU seemed to work fine. I sent him a new A/D module, but it yielded the same results. So, I asked to have the entire rack (CPU, X80 input, J71UC24-R2 serial link, Y18A output) sent in for testing- maybe the main base would be faulty. I swapped the customer's modules into a main base on a machine at our factory. I saw the same problem my customer saw. Everything worked fine with a different CPU. At that point, I believed my customer's CPU had a fault, despite a 100% program verify in Medoc. Out of curiosity, I loaded a newer version of the same program, and the "faulty" CPU worked just fine in a rack with everything but the serial link module. (Our new software version eliminated need for the serial link module.) Next, I tried another CPU that had very similar trouble with data from the A/D module several years ago. This machine's hardware and software is identical to that of our recent customer. Apparently, this unit had been hit by lightning, so I did not troubleshoot much after seeing the failure to process data from the A/D module several years ago. The CPU has been sitting without a battery for at least a year. Well, it took the new program and worked just fine! Strange? It gets better. I then reloaded the older program and tried both CPUs with the old hardware configuration and both worked correctly. Very puzzling. I'm not a programmer and wonder if anyone has any ideas why these CPUs had the same faults and now they work again. Was an internal relay somehow fixed? Memory registers cleared? I welcome all suggestions and thank you in advance.