Hmm, so if if I have this tight the evidence is that these devices all suffer some kind of hardware failure after a few months. It always manifests itself the same way, as high error rates, yet the device never quite dies completely, and there’s no obvious signs of stress?
If that understanding is right, I’d be starting to think outside the (modem) box. What else, other than modem failure, could give rise to the same symptoms, and yet always appear to be fixed by installing a new modem?
One possible thought that occurs is the modem lead. If the lead’s terminals have become degraded or dirty it might conceivably work ok-ish on a ‘factory fresh’ brand new modem with sparklingly clean terminals in the sockets and full spring tension. But then worsen over time as the modem contacts lose a tiny bit of tension, and begin to oxidise themselves. Just a thought?
Another suggestion might be along the lines of surge-arrester technology, and overlapping with discussions earlier in the thread, with B’Cat. Some surge arresters suffer progressive degradation with each arrest they make, and eventually short-circuit. I also vaguely recall that some UPS can present a waveform to connected equipment that can cause frequent ‘false arrests’ to be made, hence degradation and early death of surge arresters within that connected equipment. That really is a long shot, as I’d expect the modem to fail completely (blow a fuse) if a surge arrester on the supply shorted. But mentioned again, just as possible inspiration...