Actually I can't remember situation when people voluntarily listen (as a consumer) to solo violinist in nearfield like typical microphone (at least some of set) placement.Which raises the question, when a solo violinist is recorded and played back in studio conditions, should it sound like that violinist playing that violin in that studio, or your (romanticised?) idea of what a violin "should" sound like, probably achieved by "aggressive equalisation"?
Consumer listening assumes common reverberant space with performer and significant distance, so some sound engineer processing is mandatory for transforming raw material into commercial record unless it's puristic stereo-mic record literally "from listening chair".
Actually, we need to see harmonics content in case of real load at different levels.The only thing that changes is that higher order harmonics get more wight. But even this would not change much if we talk about 0.000x% thd of a good modern amp.
If we use even just old Shorter's weighting with (squaredN)/4 coefficients (n=7 means multiplication by 12, n=10 - to 25), long trail of harmonics is not good thing at all.
High powered amps at low level can have relatively high "classic" THD and weighting will make it much worse.
For example, recent Stereophile measurement of Mytek AMP shows extremely long trail of harmonics and I think it will have very bad numbers weighted.