I'll ask again. To what measurable characteristic do you attribute "pleasantness" of sound.
I fully agree. We basically have two common ways to describe the transfer-function of a system, time-domain and frequency domain. A DAC (including volume control, digital filter I/V- stage, analog filter and buffer) finally has a transfer-function that transforms the digital data-stream into an analog signal. We measure the final analog signal, so we measure the whole reproduction transfer-function.
Both time-domain (the signal you see on an oscilloscope) and frequency-domain description (Amplitude and Phase)
fully describe the system at the level the measurement was taken at.
We can measure the transfer-function in the frequency-domain with such a high precision and resolution, that everything audible should be covered.
This basically leaves no room for audible differences.
Of course these measurements are performed at a certain signal level and in case there would be strong level dependent nonlinearities (like we face them in speakers) , we might not really get hold of them.
Years ago we were measuring with sinewaves only and the argument was that this is a "quasi-stationary" stimulus compared to music which is much more complex. I started to use multitone for distortion measurements more than a decade ago for exactly this reason.
There once was an paper that tried to connect between measurements and subjective performance of speakers. The measurements could not predict the subjective judgement until they took measurements based on a multitone stimulus. Unfortunately I do not find this paper anymore.
Here is a comment from Linkwitz about multitone for speaker measurements:
https://www.linkwitzlab.com/frontiers.htm#G
There is one thing, that to my opinion has a large impact on our perception and this is our expectation. If we expect some unit sounds better we will really perceive this unit sounds better - our brain is fooling us. The only precaution against this is a thorough A/B blind testing.
With thorough I mean e.g. that the listening level has to be perfectly matched. 0.5db difference is too much. We would not realize "it is louder" but we would e.g. conclude "it sounds more dynamic". We once fooled us with as little as ca. 0.3dB difference; after my colleague had aligned the level to within better 0.1dB we were not able to distinguish anymore.