I think there are two changes that may have some impacts on the 48 vs 96 kHz sampling decision D+M would consider, and make in future:
1) Number crunching: That was before the availability of the MultEQ apps ($20 and $200 versions). With the apps, Audyssey filters are also calculated/created by the external devices and then uploaded to the processor, just like DL. Audyssey uses FIR only, whereas DL uses both FIR and IIR filters, as they claimed they wanted to leverage the best of both types.
FIR by nature uses more resources, as minidsp cited: "An FIR filter requires more computation time on the DSP and more memory. The DSP chip therefore needs to be more powerful." in their article on FIR vs IIR.
2) Once the number crunching is done, filters uploaded to the device, the device's DSPs still have to implement the filters and executive them to do their job, so DSP power is still, as you mentioned, a factor. However, since D+M have now upgraded their DSP chips from the ADSP-21573 to 21593 that is much more powerful, at least if you compare their DMIPs, the 21593's double that of the 21573. The same DSP chip is used across the line from the X3800H to A1H and AV10. I don't know if D+M upgraded the chip because of their anticipation for the DLBC and ART.
Regardless of their reasons for the upgrade, if they now can take advantage of the upcoming ART, they should be able to run Audyssey at up to 96 kHz. They may not do it though because they know full well the 48 kHz vs 96 kHz if just a psychological thing that has no audible effects in theory, and most users of Audyssey won't make their purchase decision on this anyway, and those who do will likely pay for the DL licenses.
Not only that the question comes up often, the same arguments back and forth, often senseless, come along with it every time too.