I agree. Unicorn has a solid state power amp around, but all the test programs generated with it thus far have been 'less than ideal' at best.
Frequency response of the loudspeaker drivers will differ with amplifier output impedance.
A solid-state poweramp designed for PA or hi-fi will have low distortion, but very low output impedance.
Solid-state guitar poweramps will have rather high output impedance, to better sound like tube amps. But they may not be designed for low distortion.
Transient response is also affected.
Low output impedance means the electrical control of the speaker is more accurate, so when current stops flowing in the voice coil, the coil stops moving faster.
High output impedance means when the current stops the speaker keeps moving. This is called overhang. It's most noticeable at the resonant frequency (around 110Hz for an OS Mesa IIRC).
This also applies when the current drops from one level to another. Basically, poweramps with high Z-out cause the speaker to store some of the energy from the signal, and release it again when the signal drops, so dynamics are smoothed out, giving a fuller but somewhat less detailed sound.
So, if you don't want to (or can't) disable poweramp simulation pre-Nebula, a clean poweramp is necessary, but it must have high Z-out or the speaker won't sound the way it does when driven by a guitar amp.
I'll add that keeping it clean by using low power levels isn't a solution. Speaker colouration can (depending on the speaker) change
lot as cone excusions increase. An accurate cab model of cones that are hardly moving wouldn't be very satisfying
![Sad :( :(](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
.