if the differences in the music are for sounds below -80db, then usual listening should never be able to discriminate between 2 dacs (or amps, or cables or anything). that's my vision of things and when I call something inaudible.
because how something is audible when playing no sound doesn't matter. we're buying audio systems to listen to music.
from my own pseudo trials of very dubious nature, I seem to be easily bothered by sounds around -60db below music. and up to -70db/-75db on silent parts while paying mighty attention and knowing what I'm looking for(all that while avoiding listening loud because I actually hear less from loud music). I'm pretty much never listening at 90 or 100db.
something at -75db with music playing above, I honestly don't hear it. and I'm not talking about white noise, I'm talking about another music or my own voice recorded at -75db and under, on music with the loudest part close to 0db. my tests were not rigorous but the scale of sounds was always in that range.
so -80db (or 0.01%) sounds to me like a very safe zone of inaudibility when music is playing at normal levels at the same time.
based on that, if all the measurements (I feel like a broken record) of time errors, distortion, and noise are below -80db. if the frequency response of both gears have less than 0.2db variations, if crosstalk is around the same values. and if the same voltage gets into the amp. then I can positively say that they will sound the same to me, and I'm pretty confident that they will sound the same for pretty much everybody.
it's not theory, it's not wild guess, if all those factors are true, I will end up with music playing at the same volume level on both gears, and sounding exactly the same up to -80db, sound level that I don't hear when playing music. that's my definition of sounding the same and a lot of dacs should be able to deliver just that without involving big prices.
so if someone is really having audible differences on 2 dacs measuring well under -80db for everything , I would say that the measures are wrong on one of the dacs (manufacturer specs are often very wrong).
and most of the times, I would guess that differences are frequency response related(sound signature) or volume level related( just like cables). and impedance might be one of the reasons why.
that is why a good DBT will show no difference. because a good DBT would try to get rid of any frequency or volume level mismatch before starting the test, to free the listeners from those bias and let them concentrate on the rest.
now if money was not a concern to me, I would probably want to buy a benchmark too. and I might when everything else in my system will be close to perfection. let's be honest, knowing I have one of, if not the best, can be very much satisfying I would guess.
but then I certainly wouldn't go for a tube amp ;) . cause spending money to get rid of -90db noises when a tube amp will create distortions probably above -60db (0.1%), it doesn't sound like a reasonable move. as always it's a matter of proportions and what we are really trying to obtain. nice sound or real sound?