Which is the most stable linear LED current regulator here?

Which of the following, A, B or C is the least likely to oscillate?
(LTspice simulation and schematic attached)

1) The gain is so low I don't see why you need loop compensation.
2) I would change the emitter resistors like this. I think the current sharing will be better.

Ron, R5 and R6 need to be changed to 20 ohms for the same LED current (or V5 to 0.5V).

I did a simulation with a 10mV base-emitter offset added to Q2. The original circuit generated a current difference between the two strings of about 6mA for that. Your circuit modification had a current offset of only 0.5mA with the same voltage offset. Even a 100mV offset generated only a 5mA current offset, so your mod is like an order of magnitude better for base-emitter offset voltages (which makes sense since the degenerative emitter resister is increased from 1 ohm to 20 ohms).

Crutschow,
Yes 20 ohms. There was two LEDs worth of current in one 10 ohm resistor. Now there is only one LED of current in each 20 ohm resistor.
The point is to get the most voltage across the stable resistor and the least across the B-E that is not stable. (temperature stable) and (voltage predictable) Thank you for fixing the error.