Q2A and Q2B are part of the exponential converter circuit.
Typically one of them does the exponentiation, and the other is used to compensate for temperature drift in the first one, so the degree of matching and the thermal contact between these will affect your tracking and offset over temperature.
What does a video oscillator gain from having an exponential response?
It lets a linear step in CV lead to a doubling in line density, which generally feels more “natural”.
It also means that if you have two osc’s set to different frequencies you can
modulate both together with the same signal while keeping the ratio between them static.
iirc, the castle clock osc does not have an expco, so if you have that you can compare.
They are absolutely less critical than in an audio module, since the ear is very good at picking up on small errors.
For what it’s worth, I’ve been thinking about trying to mod in a secondary linear input to one of my IX’s.
What happens if they’re not matched?
The tracking (i.e. how well a certain step in voltage correctly becomes a certain factor change in frequency) might change as the room or synth warms and cools. If your patches generally requires specific controlled frequencies or ratios, this will lead to you tuning a lot, while a properly compensated VCO is more a set and forget - type of deal.
Q’s 4A, 4B, 5A, 5B, are all part of the linear triangle VCO core, combining the output of the exponential stage with comparator output and driving the integrator cap.
I’m not entirely sure what bad matching here will do, but i’m guessing slight drifts and some amount of wave asymmetry.
.