The pulses of heat (post #36) probably points to intermittent silicon. TR311, TR312 appear to act as switches taking VR302 out of "play" for class AB. Problem is the VR is permanently out of play ie, for class A.
Normal TR311 E= -0.57, C=-1.14, B=-1.17 looks wrong.
Maybe E= -1.17, C=-1.14, B=-0.57 so VR302 is bypassed.
For class A you have TR311 E= -0.68, C=-1.25, B=-1.27
Maybe E= -1.27, C=-1.14, B=-0.68 so VR302 is still bypassed.
TR311 is controlled by TR312, this is possibly failing. Maybe a solder
bridge around TR312, or issue (orientation, short..) with D305 or a
problem with TR315 or that "diode/shorting link"