To wrap this up, Sequential didn't divulge what transpired (as expected), but here's what I found. Sometime in the early 2000s, Dave Smith apparently contacted Doug Curtis (of Curtis CEM chips) to develop a new version of the CEM3396 (synth-on-a-chip) for new keyboards. Dubbed the CEM3397, and PA397 specifically for Dave Smith, this new chip added stereo capabilities to the infamous CEM3396. There are two PA397 chips for each Poly Evolver voice, which seems to correspond to each pair of voices (1 & 2, and 3 & 4). Note that they don't make the PA397 chip anymore, but Behringer seems to have a clone chip available as of 2020. The chips use analog circuitry to generate oscillating waves based on various parameters (there's a datasheet available for the mono CEM3396 on the web), so the core oscillators are definitely analog - how the digital signal processor and analog to digital and digital to analog converters are combined with the oscillators is another story. For my particular problem, it's obvious that transistor Q26 was replaced, which I assume corresponds to Osc1 (left channel of the first oscillator). The replaced transistor is connected to the PA397 chip, and also the main DSP chip (which is out of production as well). Unfortunately, they don't make that particular transistor anymore (it's a MMBT3904 signal transistor), but there are a few equivalents that are close.
I also discovered another couple of issues, one of which was a cracked solder joint, likely from early (mid-2000s) versions of lead-free solder, which has been proven to degrade over time, plus my aftertouch cable developed a crack somehow (solved by some conductive epoxy).
Happily, all is well now and my Poly Evolver is back in business, but the bottom line is that not only are replacement components becoming scarce, the solder itself may be aging and cause issues, although reflowing solder is much easier than tracking down and replacing rare components.