I think I hear what you're hearing. I have sine wave wavetable at shape mod 0, wave reset on, -2 octaves. For the Osc2 test, OTA filter is around 300, resonance 0, Osc2 wave reset on, triangle shape, -2 octaves. Envelope has the fastest-possible attack.
With both oscillators, the click is occasionally just a tiny bit softer. It's hard to perceive, but I'm pretty sure it's real.
In my case, I don't notice any difference between the Osc2 behavior and the Osc3 behavior. They both have extremely small variations from time to time. (I just tried the same thing with Osc1 and heard the same slight variations.)
If you set Osc3 to LFO mode and assign it to something (cutoff, pitch, etc.) with wave reset, you can tell that wave reset is working. I'd guess that the variations come from outside the oscillators, perhaps the EG or the VCA, as has been mentioned above.
But it's such a tiny difference that I wouldn't worry about it, personally. I probably wouldn't even hear it if I wasn't listening for it. And I'm not 100% certain that I don't hear it because I'm listening for it.