Your simulation is not actually using the LM324 model, instead you are using the built-in UniversalOpAmp2 model, which is a much faster opamp than a LM324. When I resimulated with the LM324, I get overshoot and ringing (marginal stability). If I add >30nF of capacitance across the 30Ω Load, then it oscillates just as you describe.
The problem is the additional phase shift in the feedback loop caused by the Darlington, amongst other things. The faster opamp has a better phase margin, so is less likely to take off...