However, if we test and compare two items or systems and find that the
listeners can't distinguish the sound using one from using the other,
then we have evidence that they need not take assumptions that they
"sound different" seriously when commenting on the items or systems.
*Unless* some other appropriately run test shows other results in the
form of evidence
that can be assessed.

I think the difficulty here is that "listeners" is a variable and so
is "test conditions". The test conditions would be not too difficult
to replicate, but the listeners could not be easily replicated, nor
could their emotional/health states at time of testing, even if they
were.

I would hazard a guess that the quality, aural acuity and perceptual
sensitivity of a listening panel could not be easily standardised, and
since the whole experiment depends on their aural perception, I'd
forsee this as a logistical problem.

How would you suggest tackling this in logistical terms? If you
randomise the panel, this would not correspond to audiophile
listeners. Maybe you would need to randomise a sample of audiophiles
who had already been tested for good hearing. Whether you would
consider musicians and audiophiles as equivalent would, additionally,
truly set the cat among the pidgeons.