Strangely, no one but its architect has written here about the sound of the full range Melquiades, and so I thought I would say something with the benefit of several months of listening to a version of it. I say a version because though the circuit is executed essentially as in the original it is hard to know what impact incidental features of the implementation have. But then it is hard to know what impact other components of the system might have, etc, etc, we are familiar with the usual caveats.
I should also say I take it as axiomatic that playback is not a transcription of a recording but a translation; it is never the same as the original performance and it could be good for reasons that have nothing to do with the accuracy of the translation. The features of playback are correctly interpreted along the lines of the features of a musical instrument replaying the original. This means summary measures such as bandwidth, distortion, etc are neither here not there, just as they would be neither here not there if one used them to compare a Guarneri to a Stradivarius. The standard summary measures are in any event only helpful in relation to linear stationary signals, and of course what defines music is that it is aggressively non-stationary and non-linear. Nobody talks about summary measures for non-stationary signals because no such measures exist. So the objective/subjective debate in audio is mostly bullshit: we do not have adequate objective measures and never will, so the only thing that will ever matter is the quality of the subject giving you the subjective and only view anyone could possibly give. If none of this rings true to you, ignore the rest of what I say.
Now the most striking thing about the Melquiades is that it seems to handle music *intelligently*: responsively to its content. I have no idea why it should do this, but that is how it sounds. The aura of resonances that gives a musical instrument its distinctive essence is conveyed as if the pattern of distortion *adjusts* to match the pattern of resonance of the instrument itself. Someone might of course say that it is simply replicating the pattern in the input -- mere accuracy of transcription -- and it may be so, but given that amplifiers with (I would guess) substantially less distortion do not do this I suspect it is not the answer. In any event, the result is a very special capacity for tonal differentiation. It is most impressive with complex music, inevitably, for that is where tonal homogeneity is most palpably felt as a defect. And I suppose it is because that kind of music is rarely the target for an amplifier designer that we see it so rarely in the designs of others.