I don't have an LM386 model, so I took a couple of LTspice's universal op amps and set up an amp with an open loop gain of 26dB, and two poles at 500kHz. I added 50k to GND from each input, then simulated MrAl's latest circuit. I tuned the pot for minimum amplitude sustained oscillation. The pot wiper wound up at 7% rotation (from gnd), so I changed it to 10k, and added 62k in series. This got the pot wiper to 50%. Oscillation frequency was 2.83kHz, very close to the theoretical value of 2.85kHz. Third harmonic distortion was -50dB at that setting, and output amplitude was about 70mV p-p. Crank the pot up to 100%, and the amplitude goes to 1.5V p-p, and 3rd harmonic distortion is about -32dB.
One difference was the lack of DC offset, due to the model I used, so I used DC feedback for simplicity.
Right-click on the op amps in the sim to see their parameters.