An interesting experiment

"not a good match by traditional thinking"

It's more than just traditional thinking, it's "Ohms Law", every stage of amplification design is based on on it and Kirchhoff's Law, without them you'd have a mess.
And yes if it sounds good to him he may have stumbled on something that has addressed a problem elsewhere.
Typical example is the Linn Isobarik speaker of yesteryear which was a highly overdamped design (not the ideal .707Q), which worked well with amps that had very mediocre low damping factor such as the Naim 250 at lower than 20 damping factor (highish output impedance). If you drove them with an amp that had a reputation of great controlled powerful bass that had high damping factor (low output impedance) those same Isobarik's had no bass, even though the amp was regarded for it's bass performance.

Cheers George

A point that should be kept in mind is that the commonly stated 10x guideline for optimal impedance matching of line-level interfaces is commonly mistated and misinterpreted, as I see it.

That guideline says that for line-level interfaces ideally the load impedance should be 10 or more times greater than the source impedance, at the frequency for which that ratio is lowest.

That is commonly misinterpreted to mean that if the ratio is less than a factor of 10 sonics will necessarily be compromised. Which is not correct. Meeting the 10x guideline will pretty much assure that there won't be an impedance compatibility problem. But not meeting that guideline does not necessarily mean there will be a problem. It depends mainly (although not entirely) on how the two impedances VARY as a function of frequency.

If the two impedances are essentially resistive, and therefore do not vary significantly as a function of frequency, and if the load impedance is not so low that it causes a significant degradation of the performance of the output stage of the source, such as a rise in distortion, and if the high source impedance does not result in excessive interaction with cable parameters, especially capacitance, then even a 1:1 ratio would be fine. The only effect would be a small reduction in signal level, which would be compensated for with the volume control if necessary.

Most (but not all) tube preamps have a substantial rise in their output impedance at deep bass frequencies, as a result of the output coupling capacitor most of them use. That will cause a significant rolloff of deep bass response if the load impedance is not substantially greater (ideally 10 or more times greater) than the source impedance at 20 Hz. Raising the output impedance of a solid state source by means of resistors or potentiometers will not cause that impedance variation, and will not have that effect. Although if the output impedance is raised to very high levels, such as the 30K number you mentioned, there will likely be significant rolloff of the upper treble, and consequent softening of high speed transients, resulting from the interaction of that impedance with cable capacitance.

A corollary to all of this is that a high degree of consistency should not be expected between the results of these kinds of experiments among different systems. How the impedances that are involved vary as a function of frequency will be different for different components. Interconnect cable parameters will also differ from system to system, as will the lengths of those cables upon which the parameters are dependent.

Regards,
-- Al

Almarg for President 2016.

If you use a passive volume control that is 30kohm it's approx highest (worst) output impedance is about 20kohm.
This 20kohm combind with high (poor quality) cable capacitance of 100pF per foot, 300pf total for 1mt will give you a HF roll off of -3db at 27khz (-1.5db at 13.5khz).
As you can see this is cutting well into the audio band. And if the interconnect's capacitance is even higher than 100pf per foot, thing are even worse.

However if you used a 10kohm passive volume control things change for the better, the -3db HF roll off is now at 54khz (-1.5db at 27khz).

And if you use better quality interconnects like 50pf per foot capacitance the figures for a 10kohm passive are -3db at 108khz (-1.5db at 54khz)

Cheers George

Almarg,
I completely agree with you. In fact when I thought of trying this, the effect you mentioned is exactly what I was looking for. clearly if one is looking of a high fidelity setup this is not the way to go. However, sometimes playing with the equalization makes things sound better, and at the end of the day this is what counts. As I mentioned earlier, the ICE amps input is probably not 10K due to the added input buffer (single ended to differential (balanced)).
For fun I also tried a 100K series pot, with the idea of exposing the nature of the amp load and IC cable. With such a high impedance I expect the cable to have significant contribution. With such a high resistance I was basically controlling the volume with the series resistor. The effect was interesting, with rolled off highs.

Anyway, I am not going to make this my permanent setup, but it is a nice arrangement to have in the toolbox.