The structure of an EPROM is similar to SLC flash, with data stored by charge trapped on a floating gate. Over time, charge carriers tunnel out of the gate making the stored data weaker. Data retention is not infinite.
Of course not, but under normal conditions it is > 100 years.
If you are lucky, that is according to accelerated tests, and not every type. Many have ~20 year spec. EPROM corruption is not very common but isn't rare either. Don't know about this particular type but some obsolete types of those times now start t loosing their data like plague.
Whatever, I have yet to see a retention failure in a 2716, even for the ones I programmed myself 30+ years ago, without being very careful.
It's a moot point, since OP's repair theory doesn't really hold up. If the EPROMs were good, copying wouldn't help, and if they had retention errors, copying also wouldn't help
The only way it would have worked was if the EPROMs read OK in the programmer, but not in the system. I don't believe OP did any fooling around with the supply voltage while reading.
100 years? Some new flash roms will not even survive that, don't forget that the silicon processes in the 70s were far less on quality and repeatability as they now are.
Quality problems mostly lead to random failures, rather than data retention errors.
Older parts were actually specified in a much more conservative way, especially the programming algorithms. When you only have 16K cells, you don't have to worry much about programming time, so you can overprogram the hell out of it.
That said, in this particular case the eproms seem to be ok. So the problem lies somewhere else.
I don't see where there's any evidence one way or the other, but I don't see EPROM failure as especially likely. If the error message points to the ADC, surely that's the first place to look.