Did the company say where the -100dB switching noise originated from? The 1mV p-p sine clearly show the switching noise. I'm thinking it may be from the power supply. Switching power supplies for digital circuits is no big deal, usually, but in very low noise analog circuits it is a different story.
The differential output did not show this noise, (and that is why differential signalling is preferred), so it looks like the noise is ground conducted.
Finding very low noise (uV) low drop out regulators is not hard, but high frequency noise can radiate, or otherwise be hard to eliminate. Every DC/DC converter has a triangle wave with harmonics on top of the DC output. If the power draw is low, as one would expect in a DAC, then maybe linear regulators may be OK. A little heat dissipated into an aluminum case should be fine. 60Hz harmonics is easier to filter out, rectifier diodes can be selected for low noise, and resistively bypassed for less switching noise. Abrupt switching of rectifiers can excite resonances in leakage inductance in mains transformers capacitance's, and could be mitigated with suitable choke/common mode components on the AC side of the rectifier.
But it looks like they cured the problem.
NiMH batteries could alternatively be used for a low noise, low voltage supply, at least to eliminate one noise source.
If the product has rectifiers directly on the power input terminals it would be easy to feed it clean DC, like from a battery. Charging can be done while not in use.
One would hope this 32-bit converter substantially improves the paltry 16-bit Red Book source material