Thanks @mark

Do you have some examples of rotary endstops? I've searched around for this quite a bit... Also, one concern I have is regarding the recoil of those couplers. I've used them before, and my concern is that if the coil "winds" a bit at pot position 0, it may have to undwind a little bit. Isn't it possible that the unwind would put the absolute 0 position off by a couple steps?

Certainly backlash is an issue with compliant couplers, and you do have to account for it. The best solution would be to sense resistance of the pots and use that as feedback into your control algorithm, but I'm guessing that's not an option.

Anything involving measuring toque on the shaft is likely to be horribly expensive.

Control electronics that measure power required to step, alerting when power (and thus torque) rises, might help, but this would be much more controllable with DC motors with encoders (servos) than with stepper motors (which can lose steps) or rcservos (which often have poor resolution pots for their position feedback).

I've managed to get 50um repeatability with a physical endstop positioned with a grub screw, which should be more than repeatable enough for a precise control knob precision.

You can also make soft endstops using something like an IR Break Beam Sensor. Mount a slotted disc on one end of the motor shaft, IR transmitter on one side, reciever on the other. When the beam passes through the hole in the slot, you know you are inside the travel range, when the beam is broken, you know you are moving outside the range.

So, you see, there are many, many solutions to your problem, you just have to think about what you want to achieve, what accuracy you need, what space you have available, how much time you want to spend getting it right and how much you want to spend. All of these factor together to get you a solution.