R15 is still connected to the wrong rail. And 640 milli ohms is way too low, no matter where it is.
And the base of Q4 is still connected to the output node. You've just made it a more interesting spiral, but you've left it connected!
i was going to mention that.... with that short there, there is now no feedback. a 4.7k or 10k would suffice. you onle need one potentiometer in the bias circuit, as the resistor from base to emitter. the resistor from base to collector should be a fixed resistor,
the next thing you might want to do is go through and "normalize" all of the resistor and cap values to match the EIA standard values. use the E24 chart for the resistors and ceramic caps, and E6 for electrolytic caps. the table shows the base value, so when you see a 680 in the chart, that represents 6.8, 68, 680 6800, etc. depending on the multiplier. the chart lists the values from 100 to 1k. so for instance, your 400k resistor in the input stage, would not be available. the closest value from the table is 390k, and that IS available. the chart i'm describing is found here: https://www.logwell.com/tech/components/resistor_values.html
or here: https://www.pc-control.co.uk/resistor-eia.htm
germanium transistors used negative supplies, because the output devices were PNP. the switch to positive supplies came with silicon transistors and the fact that NPN output transistors were common and PNP power transistors were difficult and expensive to make in silicon. quasi-complementary output stages using pairs of NPN output devices were common until PNP complements became more readily available. at that point, amplifiers began appearing with bipolar supplies.
you still need a 10k resistor from node 14 to ground. also a 10nf resistor in series with a 10 ohm resistor from node13 to ground would help reduce the tendency to oscillate, as well as a 25uh coil in parallel with a 10 ohm resistor between node 13 and the top of C1.
2N3055/2955 pairs were used up until the mid to late 80's in a lot of guitar amps, no shame in using them. before that it was quasi-complementary output stages with pairs of 2N3055 devices. there's nothing wrong with using them if you're not too fussy about a few tenths of a percent distortion.
No.
You want node 14 to be close to the DC output voltage for it to be biased at half the supply voltage. The emitter current for Q7 is in the feedback resistor R10.
The value of R15 (the collector resistor for Q7) should be much higher than only 820 ohms to make the emitter current of Q7 low and to make the open-loop gain high.
If R10 carries the emitter current for Q7 there will always be a DC input offset. 10k from node 14 to ground would set the emitter current of Q7 to 2.5mA, eliminating offset voltage. R15 should be set to develop about 650mV from that 2.5mA, making R15 equal to 260 ohms. 270 should be close enough.
Or, if you want R15 to stay 820 ohms, then Ic of Q7 would be .650/820=793uA, and the resistor from node 14 to ground would be 30k.