You are right about offset and input source impedance. I simulated by 10k volume followed by 1k, 220p. The precedent measures were on Tina simulator, with LTspice the transistor different models gave -45mv offset. By reducing the R1 to 994 ohms it became -2mv with the volume max 10k and +2 mv with the volume min 0 ohm.
The LG, I used Tina with the bellow circuit. The LTspice gives hf on output signal if the circuit is oscillating that I get if I reduce the miller capacitors. On square wave, I have a non-resonant spike of certain duration may be depend upon the speed of the input transistors with both simulators.
The input impedance of this amp is very low in open loop, and variable with closed loop. At 1khz it is about 220k. The frequency response bellow is measured for worst case with the volume at half making total source impedance 3.5k along the 220p capacitor.