And now the cause for the post above:
When listening to the first track linked I noticed at times some compression/distorsion. Looking at the op-amp output it was clear that it was no longer able to supply (all) the current + the error current needed to correct the offset. Basically the whole "system" was latching high for a good while which prompted the long overdue modifications: disconnect the +6dB gain circuitry just to make sure I'm not tempted to use it, added a current source, changed the summing node for the servo. Some other modifications were already in place, like increased supply voltages and resistors for the current mirrors. I also changed the input pair with MAT02. Overall THD has gone down a notch (from 0.019% to 0.014%) while SNR improvement is too small to even mention. I can't tell the difference by ear alone

(and getting impecable measured results was not our objective per se anyway,...although we payed particular attention to some areas like supplies, low level performance and jitter reduction). Probably better current sources and more juggling with transistor types will bring improvements on the technical side, for now I'll keep it this way for another year or so...

I still have that BEG project waiting - or to be more optimistic, advancing on the intentional side only
So here it is without repeating the caveats: