...and the ES9038PRO hump is gone

.
The solution was almost trivial: someone reported on this forum (don't have the link, it was only casually mentioned to me) launching the hypothesis that a change in the Vcm could address the ES9038PRO "hump" issue. This turned out to be 100% true. Also noted from the ES9038PRO ESS reference board schematic that they use a Vcm=1V instead of the obvious guess of AVCC/2=1.65V.
My implementation with the ES9038PRO in mono mode, 2x4channels in parallel, then summed up, 49.9ohm as I/V resistor. You can see above the results for Vcm=1.65V. Lowering Vcm leads to shifting the "hump" towards higher input levels, to the sweet spot of Vcm=0.68V, where the hump is gone for good. Attached are the THD and THD+N measurements.
I hav no idea why lowering the Vcm eliminates the "hump", not sure if Vcm=0.68V is the sweet spot for any configuration (the ESS reference board suggests it is not, since they use Vcm=1V, but for 8 independent channels) but mostly I have no idea if the same almost trivial method would apply to the ES9038Q2M. If anybody has any ideas how/why Vcm affects so strongly the distortion profile, I am all ears.
-112dB THD+N and -123dB THD, this is one good DAC

A shot of my ES9038PRO test board (one channel only installed, shown in a digital loop with my own ADC implementation) is also attached below.