vinyl can be cut, played back with content above 22 KHz - not at full amplitude, not with all cartridges, stylus profiles, the high frequency won't be even across the disc, inner grooves are moving slower, the features just become too fine, new cartridges, vinyl compounding was required by CD4 encoding which used a subcarrier for FM that reached 45 KHz
cutting heads for mastering vinyl have 2nd order roll off ~ 50 KHz, there was some experimentation with half speed mastering - but that can't be used with direct to disc
even direct to disc is limited by microphones, 50 KHz is the fastest sold for recording studio use, the overwhelming majority in use have less than 25 KHz corner frequencies, classic vocal mics may have 12-15 KHz roll off
if you want to be conservative I would use 96 or 192 to record vinyl, 24 bits is handy even if you don't get more than ~ 20 bit resolution with today's electronics - for the headroom - not because vinyl ever approaches even 96 dB/16 bit S/N
controlled listening tests can't prove that "no one can hear X" but there are large tests which didn't find the positive result that any participant could reliably identify content that was recorded with high frequencies in high res formats from the same material properly reduced to 44/16
a few published tests which do claim to show over 20 KHz discrimination with music signals have been shown to have technical flaws or haven't been replicated by other groups and accepted by the psychoacoustic Science community as proven yet - it is still an open question
noise shaped dither gives effective resolution of >110 dB to our ears with 44/16
as has been repeatedly pointed out the most practical real test is to digitize the output of your phono preamp (observing proper gain structure, not overloading the ADC) and do a live bypass test of direct vs good quality ADC-DAC with careful level matching - until you do that your opinion is suspect on many grounds
people arguing for some magical purity of analog simply do not understand signal theory - it is fair to say 44/16 may be "too close" to our hearing limits and may fall short at some points, it is wrong to claim that digital representations of amplitude, bandwidth limited, thermal noise limited signals cannot be accurately represented digitally