OK, I agree. The first thing that has to be clarified is your use of "slightly" higher bandwidth for analog than RBCD.
That slightly is as vague and as multi meaning as this :
Since digital camp does insist on accurate description of each and every bit and byte, I will counter with the undeniable fact :
ANALOG RECORD FREQUENCY RESPONSE IS NOT UNIFORMLY AND UNIVERSALLY DEFINED
What does it mean ? It means one can (hopefully not... ) play some super cut analog record with recorded signal past 100 kHz ( can be done, has been done ) with a $ 0.99 ceramic cartridge tracking at 10 gram vertical tracking force - getting a ragged response up to about 10 kHz and ruining an expensive record for good in a single play - OR can use a super cartridge with response past 120 kHz ( can be done, has been done ). These two are the maximum opposite extremes possible - with most real world equipment performing somewhere between those two extremes, preferably in the direction of the 120 khz cartridge.
Fast forward to 2019 - almost ANY moving coil phono cartridge in the market for sale today has frequency range AT VERY LEAST TO 50 kHz.
It is debatable how linear or with which deviation(s) from absolutely dead flat up to 50 khz, but cartridges that have survived in the market up to the present day can not have too wild excursions either way, up to, including and beyond 50 kHz - or the competition simply pinches them out of the contention in a rather short period.
So - IF we assume only the majority of MC cartridges available, FLAT response of digital whatever is required at least to 50 kHz in order to have a chance of even approaching their full capability. That clearly rules out RBCD, DAT ( 48k) and 96/24. Brick wall filtering and attendant problems you have so eloquently described for RBCD hold equally true at higher sampling rates - and 96/24 can not do justice to most MC cartridges available in stores today.
That leaves us with 192/24 as the first PCM capable of roughly meeting the FLAT frequency response capable of covering the frequency range of most MC cartridges today. And, since you ARE familiar with the quantization noise etc creeping up in level above certain frequency for any real world ADC/DAC, that means at least 384/24 capable "digitis" is required IF the noise above > 20 kHz is not to exceed certain limit.
DSD compounds the problem - because it is never filtered out so steeply as PCM, its frequency response starts SLOWLY AND GRADUALLY to fall off at much lower ferequencies than brickwall filtered PCM. Likewise, ultrasonic noise becomes reasonably manageable only with DSD256 - with DSD128 being an absolute minimum in this regard.
SACD - or more precisely, DSD64 - should have never become available ; it is because DSD64 that DSD in general has gotten bad rep in some circles..
I would like to stress that I used here entirely plausible and commercially available equipment that not only millionaires can afford; I did no go for the extreme final capability of analog record at 120+ khz - which will most likely remain as the record in its category, both for the ability to place it on record as well as to be able to play it back.
From this perspective, 96/24 PCM can still be regarded as kindergarten level - OK, make that primary school.
Wild excursions up to 20khz., okay; but wild excursions from 20khz. to 50khz. and beyond, is not going to determine a cartridge's success in the market.