Interesting question, let me think about it because I think there is ground for something interesting and also simple.
My motivation for this suggestion is that if we think with traditional “musical” terms, it’s not the same as how scales are considered if the instrument is a Bass, or tener, or Soprano, etc. A bass will make whole “band”/system way mora stable if it only plays root notes (then as is goes higher It can make 4ths, 5ths, and then as it goes higher one can consider "upper structures’’, for example having A-C-E as “A minor” in the upper structure, and C mayor (just C) in the Bass.
This is only the simplest example I can think of to illustrate the point. One then can later program a whole quantized musical scale in these terms, and have a narrow, yet simple, harmonically rich patch happening. In the way it is right now, If we use marbles in a wide range on notes, (so it covers, bass, medium and high), then some notes will appear randomly in the bass register wich change the whole harmony of the patch, when the bass plays an “A” then the whole patch becomes “A minor”. I propose that the system stays in C major while higher notes can play “A`s”.
Maybe I’m already repeating myself, I hope I’m clear.
I’ve been doing something “similar” to this, modulating the STEPS, BIas and SPREAD all together, so sometimes BIAS is low, STEPS allow only root or root and fifths, and SPREAD is tigh. This way I only allow bass notes to be steady notes. Then the modulation source (usually a slow LFO) allows STEP to allow other notes, and BIAS also allows higher notes, and SPREAD may or may not be higher also.
This is tremendously complicated, considering that the STEPS knob only uses half of the knob range (from 12 to 6 o´clock), so I usually use 3 attenuators for dialing the exact CV range.
The “quantization per octave” would make patches way more simple and harmonious.
The whole theory can be thought in 2 ways:
1- Upper structures. Having different chords in each register, yet still remain diatonic (or not).
2- Extended scales/ Super Ultra Hyper Mega Meta Scales (contemporary popularized by Jacob Collier, here is an explanation: Super Ultra Hyper Mega Meta Scales - YouTube ).
For the question of “what does the range selector do?”
One thing that comes to mind is that maybe it doesn’t need to affect anything to quantized outputs. Since programming scales means that the user knows exactly what one is doing (not just noodling around), then If one wants a limited range, then we just enter the notes we want.
Another option would be to “remove” the notes that are out of range. So -5/+5 would include the whole range from A0 to C8. The 0/+2 will only allow form C3 to C5 ( ? ) and 0/+5 will allow only C3 to C8.
All this while having unquantized outputs just be -5/+5, 0/+2 and 0/+5.
Thanks so much for consideration.