Hi Tom.  I don't agree with your statement "frequencies over 20k are
mostly trouble and should thus be filtered (which is the long-time thinking
in most audio circles)"  At least not from an equipment manufacturer's
standpoint.  I suppose you are speaking about if from a mastering
perspective.  But from an equipment standpoint, good high frequency
bandwidth, and the lower phase shift that accompanies it, has been a virtue
as long as I can remember (API, Neve, De Medio in the pro realm, and Harman
Kardon (Ultrawidebandwith) in the consumer realm.  Those guys clearly
believed that there was a sonic benefit to preserving frequencies above 20

I remember when CDs came out and one of the claimed  benefits was wider
frequency response.  My Pioneer RT-707 1/4" deck went out to 28kHz, so I
knew that claim was false.  I recently re-built some Ampex 440s and ran
them on an Audio Precision and was surprised how far they extended (past
30K if I recall) when they are properly tweaked and with modern tape
formulations.  And many phono cartridges go way beyond 20K - I seem to
remember some going out to maybe 40K in order to reproduce the 30 kHz high
frequency subcarrier for the CD-4 quadraphonic decoder.

And also this statement "input and output transformers acted essentially as
band-pass filters, unable to pass ultra-sonic frequencies."  There are
plenty of good transformerful (as opposed to transformerless) mic pres that
go out to 80 or 100K.  The 440s had transformers in them too.  I think
"unable to pass ultrasonic frequencies" is a bit of an exaggeration.  A
good audio transformer can definitely pass ultrasonic frequencies.  With
early digital I think it was more about the brickwall cutoff slope (and the
horrendous phase shift that accompanies it, especially using the multi-pole
analog filters of the day).  Analog equipment (and transducers - such as
phono cartridges and tape heads, which are also really transformers) tend
to roll off more gently.

As far as musical instruments not producing frequencies above 20K, that is
patently false.  It doesn't even make sense from a logical standpoint.  Why
would a physical object stop vibrating above 20 kHz just because our
hearing cuts off there?  The highest frequency emanated by an object is
defined purely by physics, and there is no magical 20 kHz limit.  Hit a
triangle and put a good mic and mic pre on it and look at it on a spectrum
analyzer.  The harmonics go on practically forever - they just decrease in
amplitude until they're down in the noise floor, way up in the ultrasonic
range.  Same for cymbals and muted trumpets.

As to whether any of this stuff is directly audible, that is up for debate
I suppose, though there seems to be plenty of anecdotal evidence to lead me
to believe that we don't completely understand the ear-brain mechanism.
I'm not sure the Geoff Emerick example is a good one though - true, an
unterminated output transformer will ring at ultrasonic frequencies, but
the Q may be wide enough for there to be significant HF boost and phase
shift well within the accepted audible range.