I believe the non 40-bit parts of the firmware are incorporated in the 272, but the noise reduction from decreased round-off error obviously aren't.
My beliefs were wrong. I originally misread the ADSP 21369 data sheet, seeing only the 32 bit fixed point precision. In fact, as Naim pointed out to me this morning, that chip also uses 40-bit floating point precision, just like the ADSP 21489 in other products. Apologies for sowing confusion.
Naim's full reply to my query about the ADSP in the 272:
"We use the same 40 bit floating point algorithm in all our SHARC based products. The algorithm oversamples incoming digital audio to either 705.6kHz or 768kHz and buffers the audio for subsequent jitter removal.
Both the ADSP 21489 and 21369 DSPs can run in either 32 fixed point or 40 bit floating point; we only use 40 bit floating point.
The 489 is a little newer than the 369 and runs a bit cooler for the same rate of math otherwise the digital audio is absolutely identical. We differentiate the sound quality using the DACs, PSUs, analogue components and mechanical treatment such as suspension."
Given all this, I can't imagine the streaming section of the 272 is any different from that on the Nova. The power supply, analog (including DAC), and mechanical treatments are all supposedly better. This explains their recommendation of the 272 over the Nova thoroughly. I'm sure the Nova punches above it's weight, but the 272/300 combo must be better. Especially with a later addition of the XPS so many others have recommended.
Have reached out to Oxford Audio Consultants, let's see if I can verify my assumptions during my next trip!