I agree that component tolerances is a factor.
Imprecision is to be expected on a 33 years old Oberheim Matrix-6 (which uses 6 Curtis 3396 chips almost identical as the ones in the REV2), because of the limited digital technology of the day (slower processors, limited resolution ADC/DAC, limited amount of memory that was expensive then). But on a ultra-modern synth, with today's extremely powerful processors, inexpensive memory, and high resolution converters, I would have expected at least 255 values for all the potentiometers/parameters, and a calibration routine capable of enough resolution to cancel out any unwanted component tolerances variations from voice to voice
on critical parameters such as wave shape/waveform duty cycle. When I dial-in 50% pulse width in order to get a square wave, I don't want 52% or 48% on some of the voices.
Would anyone settle for an imprecise and varying sawtooth rise time between voices ? I doubt it. Some parameters are critical, others can vary and offer dynamic changes pleasing to the ear. And current technology allow a precise enough control to get all of that simultaneously.
As proof of that, the extremely precise filter cutoff frequency control on the REV2 allows quasi-perfect keyboard tracking across a very wide range and on all 16 voices, which was impossible to get on any 33 years old synth, no matter how well a calibration could be done on it.
All that being said, maybe some of my critical parameters aren't as critical for some people ? Probably. But the 50% duty cycle one is as critical for awelchel (the thread starter) as for me.