You are forward biasing a transistor in your feedback loop. That is possible, but it does require some care and attention to prevent hinky oscillations, something like those oscillations might be a cause for dead op-amps.
Another cause could be current drain at the output in case of sudden input voltage increases, though I'd expect that to be of such short duration that it wouldn't cost you many op-amps in a single go.
A reason for Q2 to break would be high dissipation, 6.8V across 100ohm is 68mA. if your supply really is 20-ish volts, that means the transistor is "eating" about 13.2V, because, where else does it go? That means it dissipates 13.2V * 68mA = 890mW.
If it's a normal TO92 type that'll very likely kill it, even soldered into a PCB, when working in a breadboard (are you using one?), it's anybody's guess. (Because the heatflow in a soldered situation can be predicted, in a breadboard that's significantly harder).
As a side interest, if you've destroyed one part, you might want to change the measurement points and types a little and try again, but when you've destroyed two parts in the same set-up, that's usually the point to critically examine the construction and/or schematic.
Q1 could be having problems with base currents in case of oscillation or spikes on the input, but with a 741 it seems that one is reasonably safe.