Well, I had a different thinking of “done right”… I was refering to reliable design.
No amount of features is going to sell an analog polyphonic if it isn’t reliable, and the vintage market is littered with unreliable ones.
A very common achilles heel is connectors. If the contacts are not gold-plated then they will not be reliable for carrying power rails and critical CVs, which translates to tuning problems. Connectors with gold-plated contacts aren’t cheap though, thus many vintage polyphonic designs settled for tin-plated contacts. The problem with tin plating is it will oxidize to the point that they can’t be cleaned at all, and you lose voltage over that ozidation layer. OK for logic levels but bad for CVs and power rails. Those were issues I identified and corrected in my Memorymoog and am now applying the same corrections to my OB-X.
So if the design decision for a new analog polyphonic was made to go with tin-plated connectors, then the connectors should carry ONLY logic level signals. This necessitates putting the DACs and S&H buffers on the voicecards to prevent analog CVs from being propogated via tin-plated contacts. So many hybrid analogs got this wrong by piling logic and analog signals onto tin-plated contacts. You can get around gold-plated contacts for power rails by using local voltage regulators - an approach favored by Oberheim, but the gain was erased with analog CVs on tin-plated contacts and no thermocouple compound between the tempco resistors and the CA3086 array performing the expo conversion (insert facepalm here).
Local DAC/S&Hs does carry a higher price due to redundant circuitry on each voicecard, but it does permit multitimbrality and voice modulation. Global modulation sources such as pitch bend and LFOs can be done in software and can modulate via the digital domain, which has been the approach since the 1990s.
As I said, there’s a right way and a wrong way to design a polyphonic.