I haven't taken the time to read most of this thread, but I believe I can provide some insight into the points raised in Rebbi's post just above.
As you no doubt realize, global feedback affects the characteristics of an amplifier in a number of different ways. Everything else being equal, an increase in global feedback will reduce gain (as you observed), output impedance, and total harmonic distortion. It will, to some degree that may or may not be objectionable, increase transient intermodulation distortion, and certain odd-order components of the harmonic distortion to which the ear is particularly sensitive.
I suspect that the reason for the changed perspective you described was not due to the effects of feedback on the sonics of the amplifier in itself, but rather to the interaction of the amplifier's differing output impedances, at the different feedback settings, with the impedance characteristics of the particular speakers. Many speakers, including many two-way monitors such as your Reference 3A MM de Capo iA, have substantial impedance rises occurring somewhere in the mid-range. For example, see
this impedance plot for the de Capo i, which rises to more than 10 ohms at around 2.5 kHz, while being close to 6 ohms in the mid-bass and much of the treble region.
Given that kind of speaker impedance characteristic, and everything else being equal, higher amplifier output impedance (resulting from less feedback) will result in at least a slight increase in emphasis of frequencies around 2.5 kHz, relative to higher and lower frequencies. While lower amplifier output impedance (resulting from more feedback) will not produce that emphasis, at least to the same degree. And from a subjective standpoint greater emphasis of the mid-range and lower treble is typically perceived as a more "forward" presentation.
Regarding your concluding question, about how your observations may extrapolate to a SET having no global feedback, I would be hesitant to draw any generalizations. Different zero feedback SET designs may vary dramatically in their output impedances (as well as in many other ways, of course), and will vary in the interactions of those impedances with the impedance characteristics of whatever speakers are being used.
In an ideal world an amplifier would be able to achieve zero distortion (of all kinds) using no feedback, and it would have an output impedance whose match to the impedance characteristics of the speaker is consistent with the overall voicing of the speaker that was intended by the designer. But of course in the real world tradeoffs always come into play, and the optimal amount of feedback will inevitably be dependent on some combination of the particular amplifier, speaker, system, room, recording, and listener.
On another note, I've been remiss in not yet extending congratulations to Brownsfan on his retirement. It's certainly proven to be a great institution in my (and my wife's) experience :-) And of course best of luck with the new locale.
Best regards,
-- Al