[Mondial]

Hmmm. When you checked with the scope, what did the distortion look like with the speaker connected? Was it on only the top or bottom of the waveform, or did it distort equally on both?

Feedback amps are really hard to troubleshoot because a problem anywhere in the circuit will affect the signal from input stage to output. Sometimes the only way to fix something like this is to check each part individually, rather than try to trace the problem to an individual stage while operating.

Another complication is that it appears that the voltage measurements on the schematic were made with a sine wave signal applied and load connected. There is no way you could have .4 V across a .47 ohm emitter resistor from idling bias alone. That would be .4/.47 or 850 mA ! At that point the output transistors would be dissipating 35 Watts and cooking. Rather, under no signal conditions you probably should see 10 to 20 mV across R95 and R97.

I guess I would at this point remove the remaining transistors and test them. While they are out of the circuit you could check most of the remaining passive components without removing them, since most will be isolated with the semiconductors removed.