Electronic – How to pull an op-amp output low without transients

I am designing a simple 'junk box parts' adjustable constant-current dummy load. The following schematic seems to be the most promising design, where R4 and R3 are part of a current adjust potentiometer.

I would like to add overtemperature protection to the heatsink of Q1, and have the detection part of that sorted (LM35 sensor, voltage reference, and another op-amp in the LM324 as a comparator with some hysteresis). My question is how I can best use the output of this op-amp comparator to switch off current through the darlington pair Q1/Q2. The possibilities that have occurred to me are:

1. Use the comparator output to drive the base of an NPN or NMOS connected between the base of Q2 and ground, sinking current away from Q2 when the comparator output is high. The disadvantage of this is that, at least in LTspice simulations, large transients occur during switching (resulting in large current spikes through Q1), as the output of U1 switches between being railed high and being at whatever its equilibrium state is when feedback is active.
2. Use the comparator output to drive the base of an NPN or NMOS connected between R5 and ground. The disadvantage here is that, while there are no switching transients, the output of U1 is not being pulled all the way to ground when turned off, resulting in leakage through Q1/Q2 of several mA.
3. Replace the direct connection between the current shunt R1/R2 and the inverting input of U1 with a resistor (10K-100K) and use the comparator output to drive a PNP or PMOS that pulls the inverting input of U1 up to (nearly) Vcc when the comparator output is low. This seems to me like the most sensible solution, except that my junk box has no suitable PMOS parts, and using a PNP is a little awkward with an LM324: the output of the LM324 cannot pull higher than Vcc-1.5V, so the emitter of the PNP will need to be two or three diode drops below Vcc.

Are there straightforward ways to mitigate the disadvantages of solution 1 or 2, or would I be advised to go with solution 3? Or are there any other good solutions?

(I should mention that I ask this question as a hobbyist who is much more interested in learning about circuit design than in making this particular dummy load. I am aware that the best solution is probably to just use a big enough heatsink in the first place, and something like a BTS141 (MOSFET with built-in overtemp protection) instead of Q1/Q2.)