It is essential to capture everything on the tape so that speed correction and distortion reduction can be performed. That's the fact of it - and it makes the discussion of audibility of 96 vs 192 moot.

You can't hear 100kHz or 152kHz (typical pre - ATR Ampex bias) but the results of capturing it and tracking it are extremely good, and beyond debate, at this point. 

Once captured to 384 and processed do as you wish, do your declicking or editing 
We also have a multiplex strategy for folding it to 24kHz for those who digitize direct at 96 that works well, but it requires our hardware.
Meanwhile however - drive storage space is cheap and at the higher rate through some heads and converters the time-base correction could be done whenever it was desired, often works from existing transfers played on 102s, which reach to about 120kHz stock, albeit at much lower S/n levels than ours. But still good enough to work wonders with the sound. 

Working on a 1963 Abbey Road tape we found plenty of 80kHz raw signal when we transferred at 192/24 from an Abbey Road C37 Studer - pretty much their go-to machines (and J36) until 1967. 
This confirmed that it would be relatively simple to dewow and deflutter every blessed Beatles recording from Love Me Do through Sergeant Pepper, using the existing digitizations at 192/24 if they existed.
But best as we know they don't, since Abbey Road standardized and defends 96/24. So until they pull the tapes again and do retransfers it's impossible, since the existing useful signal is lopped off.  But to have that happen you have to get past the policy which is way harder than creating the technique. And get past this discussion into a "best practices" discussion about capturing the mechanical metadata of the recording. Which is easy. 

Dan's paper as far as it goes makes sense, particularly back in the days when ADC settling time and clocking error and other flaws were less apparent at lower Fs... The converters worked simply did their job better at 96... No claims were made about utility or audibility, that's not his thing. From an engineering POV back then he was right...  Pretty sure 192 and higher rate low bit over -sampling  converters like the Pyramix 12.8mHz/5bit Horus (which sounds phenomenal) have caught up in the years since Dan wrote that paper. 

Please pardon the misspellings and occassional insane word substitution I'm on an iPhone

> On Aug 29, 2014, at 11:33 AM, Lou Judson <[log in to unmask]> wrote:
> I appreciate your logic, but *I* feel upsampling does not give as much benefit as converting at the high rate. Delivering upsampled files claiming they are original transfers amounts to deception in my book. "… no one will know the difference" is the edge of a slippery slope ethically.
> Higher rates give better cleanup and processing for me, but I also feel 96k is good enough for anything... Agree with Lavry on that...
> Thanks for the article link!
> <L>
> Lou Judson
> Intuitive Audio
> 415-883-2689
>> On Aug 29, 2014, at 8:12 AM, Rob Poretti - Cube-Tec wrote:
>> If ABX listening is irrelevant from a business standpoint, then I would suggest simply up-sampling at the last delivery stage... no one will know the difference right?  Why waste time and money?  ... and you can still tout it as 24/192.