Today we tried some careful tweaks at the pots.
VR4 is maybe for the switching of the higher two rails (class H), VR3 is for the adjustment of the bias, in german "working point?".
First of all we checked the voltage on R146 (between the two BJTs that are driving the final stage. The good channel has 6.5V, the bad one has 5.4V across that resistor. So I tried tweaking that with VR3, but nothing changed much when adjusting it. +-0.1V, thats it.
Now we analized the circuit from front to back....
Q3 and Q4 are the differential inputs, Q2 is a current source, that delivers 1.5mA. This current divides to the two ways of the diff amp input, so at idle state 4,5V are at the resistors R62 and R63 (below the diff amp). Measurements tell the same and are equal on both channels. Tweaking VR1 changed the offset a little bit but gets back to its 1V in <1 second (it seems there is some capacitors that charged up or something before returning to the faulty level.
Q21 and Q22 (right of the current source) seems to be a current mirror circuit if i remember correctly from school some...SOME... SOME years ago.
Now there are only Q26 and Q27 left, they seem to do some voltage amplification but sharing a pair of feedback resistors....? weird.
Now we stopped exploring and maybe there are some more hints to check...