The Bosstone circuit presented at the start of this thread implements a typical shunt-shunt global feedback configuration over Q1 and Q2, where the feedback network (560k + 0.022uF + 560k) moves from closed to open loop along with growing frequency through C2. The schematic below shows my simplified analysis version, where the simulation model is on the left side and the small-signal model on the right side:

From the basic feedback theory, the gain for the shown circuit in closed loop (low frequency limit) is (RB1a+RB1b)/RS = (560k+560k)/10k => 40 dB. In open loop (high frequency limit) the total gain is determined by Q1 as -gm1*RL, where gm1 is the transconductance of Q1 and RL is the effective load on the collector of Q1. Since the input impedance of the "emitter follower buffer Q2" is huge, the RL in open loop is approximately 560k (RB1b), since the feedback loop is split into input and output sides via grounded C2 and RB1b appears parallel to the huge buffer impedance. Here is my simulation on the simplified circuit, showing the frequency response of Vout/Vin:

Note that the second 18k resistor affects mainly on the biasing of Q1, it does not have much significance on the gain as it is parallel to rpi2 in the small signal model. Similarly the effect of the "magical 50pF" cap is lost, because the BJT internal capacitances are most likely higher that this, resulting in a natural HF roll-off.

If someone is experiencing oscillations in this circuit, I would personally try to add a small (maybe 220 ohm) emitter resistor to Q1. This will add local negative feedback, and hopefully stabilize the high gain response of Q1.

I see the Q2 only as a buffer in this circuit. For the loop gain in closed loop state it contributes as a significant gain multiplier hfe2+1, and in the open loop it presents a high-impedance load to Q1.