I did not catch what paper you quoted? And what the heck does "heavily oversampled" mean?
In any event, oversampling DS modulators do indeed have high-amplitude ultrasonic noise and a noise floor that rises with frequency; that's what they are designed to do! They shape the in-band noise out of band, into the HF regions, then filter it off. That has nothing to do with jitter. I have a feeling I am missing the context of the paper.
The theory was presented (loosely) in one of my fundamental jitter threads, and says that clock rate does not matter for jitter, just resolution (number of bits) and signal frequency. I do agree wider bandwidth may lead to greater noise, but that is not the same as increased jitter on the clock. Higher clock frequencies may have higher jitter, or not; most are in a controlled loop (DLL/PLL) so it is the loop bandwidth that sets the jitter, not the carrier frequency (to first order).
HTH - Don
p.s. How did we get here from the bit rate needed to sample vinyl? I wasn't intending to post here as I don't really know -- too much depends upon the listener and what bandwidth is really needed, which is pretty subjective. Alas, I suspect very high rates are needed to capture the full bandwidth of the ticks and pops!
(It's a joke, guys, don't crucify me over it...)