Add my thanks to Marvin for the photo. I saw the summer IC, pulled up the schematics and there it was. No wonder the (not totally read) schematic wasn't making sense. I'm not familiar with DAC circuits, so I also must ask, is it normal for the U15/U17 opamps to be biased with such an offset?
It's interesting biasing indeed. On my over-simplified-schematic had it marked as 3.3V but that's the AVCC_R rail, in reality it goes thru a resistor divider with 10k/3.3k ohm so the bias on the non-inverting input of RT6862D is 0.82V. The output from ES9038Q2M is a differential signal with zero at AVCC/2 so that is 3.3/2=1.65V with full scale out being 0.906xAVCC peak to peak. So that is 0.906x3.3=2.99V (let's say 3V) peak to peak meaning full output from the DAC is centered at 1.65V with peak at 1.65+1.5=3.15V and bottom at 1.65-1.5=0.15V. And that signal is then biased up 0.82V at RT6862D. Just a note that the opamps are running at +-5.5V provided by R1286K010F (datasheet https://www.mouser.com/datasheet/2/792/r1286-ea-1076068.pdf).
So I guess they wanted a higher balanced signal to convert to single ended with RT6863?