Very interesting thread, this.
On the subject of the audibility of the top octave, it seems to be discussed as if we could turn that top octave on or off, like the cut off rate were a rectangular filter response. In practice, that's not how it goes. For example, in analog recording systems, affecting the response at 15KHz also affects 10KHz, and to a much greater extent, 20KHz. In most cases the loss of response in the top octave is fairly broad. The area of non-flat response often extends well below the top octave, though to a lesser extent. How audible the total system is relates to the rate of response change and total area of response affected. Those gentle 6dB/Octave roll-offs that result in, say -6dB at 20KHz will also be down -2.4dB at 10KHz, 1dB @ 6KHz, and that's audible. The differences between a steep low pass filter, say 48dB/oct. at 15KHz and the same one at 20KHz may be harder to hear though.
For test observations to be valid, we'd need to know the exact filter characteristics used. Just saying you tried an EQ and could/couldn't hear it isn't really meaningful. We'd need to know what kind of EQ (peak/dip, cut, shelf), what bandwidth/Q the specific filter had, how much attenuation, what frequency it was centered at, etc. Otherwise we're comparing very different fruit.
We could also discuss average hearing loss as a function of age, as adult males loose the top octave first, and wouldn't notice its absence, where a 20 year old who's never been to a rock concert (yea, try to find one of them!) and has undamaged ears might hear up to 22KHz. Yes, I did, and I miss those days, barely 12KHz now.