Google did find s
There's an interesting op-amp on each channel that is sand-off, not sure what it is for.
I'd think that's responsible for the power amp input stage. Looks to be a dual opamp, and I'm pretty sure I've seen that face before... if I had to place my bets I'd say that's an RC4580I. Used as poor man's fully differential opamp, I guess? (Would explain why what little distortion is there unloaded is domanent H3.) What follows must be inverting, as the capacitor as between pins 2 and 6, the inverting inputs. Either way, the opamp is clearly not the bottleneck, the output stage seems to have some current-related symmetry problems instead (
distortion mechanisms #5-7 may be worth checking for).