To be honest, most people I have encountered who call themselfs audiophiles doesn't seem to be very interested scientifical measurments, and technological innovations that can actually be proven to work. So in general basis, I'm prety glad "audiophiles" doesn't deside the desgins. (However there are of course audiophiles that actually know and follow the science behind sound).
Having sympathies in both camps - and engineer who values the sound of my music - I'd defend the audiophiles in that many engineers gleefully laugh off their objections and comparisons between CD players, digital links, DACs etc., saying it's digital therefore perfect. In reality there are real
problems with all those things. Pretty much few or no CD readers extract the data reliably bit-perfectly, and few or no digital links transmit it reliably bit-perfectly either.
Also, many engineers love to show frequency response graphs like they're the be-all and end-all of audio performance, when they're usually generated with fairly static generated signals that have none of the variation in the temporal dimension that real music has, so they don't measure the temporal responsiveness of the system. I think (?) that they're usually generated with a sine wave too, so don't measure the ability to render multiple frequencies simultaneously.
So, I'm not too surprised that many audiophiles reject everything but listening tests, given how the engineers' techniques and measurements have so often failed them. Personally, I think technical tests are the best for the digital realm, and once you go audio, the ear is the primary measurement. That's why I'm so frustrated that none of the digital audio interconnect techniques are technically reliable.
The problem is that many audiophiles do reject listening tests also. When differences can't be proven with double-blind-testing claim that thats not a good way to test.
And to be honest, although bit-errors are annoying, I have never seen any listening tests where a difference has beens shown (if the faults doesn't manifest themself in obious pops and so on, in other words the faults seem to be either ver obivios, or not noticable). Do you have any more data on this? It would be pretty interesting.
Frequency-responses are created in a number of different ways (if you are measuring a speaker), but a sweeping sine is one of the more normal ways.
The ear is a pretty magnificent organ, but that doesn't mean that we shouldn't attack the problems with science.
But talking about technicall reliable, the squeezebox should be pretty reliable? Transfering the data through a regular network should be pretty safe, data-integrity-wise.