Hi Chris,
I think that (Klipsch) horn speakers benefit from transconductance amplifiers.
But, I think that amplifiers with an output impedance of about 40 Ω are not realy transconductance amplifiers (or current-drive amplifiers). Those are more power-drive amplifiers, with an output impedance around the impedance of the loudspeaker (within a factor of 10). So, amplifiers without feedback and valve amplifiers are all (more or less) power-driven amplifiers.
A real transconductance or current-drive amplifier has an output impendance of 400, 4k, 40k or more Ω.
You read and mentioned already the paper of Esa Meriläinen and learnt that in the small-signal domain of drivers there is a huge benefit of current-drive. I don't know whether the Klipsch drivers have non-conducting voice coil formers (bobbins). If they have, the benefit is even greater (as you could read in his paper).
I think in the large-signal domain around resonance a horn system is probably inherently better than other systems. But because of the elimination of electrical damping (one of the non-linear distorting effects of voltage-drive) you have to cope with the rising response around resonance (with -real- transconductance Qts equals Qms of the driver!). EQ does the job. You also need EQ for the rising response on the high-end due to the elimination of inductance effects (damping and non-linearities).
But if a horn is used (far) below resonance there is a drawback. Around resonance a driver is sensitive for second harmonics of frequencies half of the resonance and for third harmonics of frequencies a third of the resonance (so, with driver resonance at 60 Hz, the driver is sensitive for second harmonics of 30 Hz signals and for third harmonics of 20 Hz signals). This goes for every driver, but with a closed system you can realize enough acoustic damping of the resonance and of the sensitivity. I don't see how you could do that with a horn system. On the other hand, I do wonder whether a horn is ever used so far below resonance...
So, I think that every driver thats work with an electromotor (F = Bl · I) benefits from transconductance / current-drive in terms of distortion, bandwidth and stability (no destabilizing effects of heating voice coils). (I don't know how piezo-drivers, electrostats and other exotics react to current-drive.)
But... But traditional crossover don't work with transconductance! They need a redesign or elimination. Esa explains how to build passive crossovers for current-drive. Even better is to use an active configuration. You can use DSP for driver correction, time alignment of drivers, crossover, time alignment of loudspeakers and for room correction.
Leaves the problem of availability of (real) current-drive or transconductance amplifiers. I build one myself (it is no rocket science, just changing the feedback, and the book of Esa helps), build my own crossoverless 2-way system (sorry, no horns) and bought a 2x4 channel DSP system with DAC. The sound is crystal clear, soundstage is great and stable and an active system is very efficient. I have 4x 80 W available, but I wonder if I every used 1 Watt per channel...
Send me some Klipsch-horns and I will tell you how they sound. :-)
How do you go on your quest for lower distortion and better sound?