Valves are not linear. If you 'push-pull' in Class A then the non-linearities cancel out to some extent. You can then use negative feedback to tidy things further.
A SET usually has no feedback and no mutual cancellation. Hence more distortion.
I think.
Paul
My 2p, FWIW, IMO, YMMV, etc:
Valves (especially triodes) are certainly more linear than bi-polar transistors. That's the point of using them - they're
just about linear enough to use without the crutch of feedback. I'm of the opinion that the better the open-loop linearity, and the less compensation (feedback) you need to use, the better. I believe this brings benefits when used appropriately, i.e. with high-sensitivity speakers and a benign impedance curve to minimise the problems (which are real). Of course they're going to perform poorly when partnered inappropriately. Not all speakers require the lowest damping factor possible - for some (e.g. old Tannoys, having a slightly higher Zout can help in the bass).
In this context SE vs PP in not relevant - the output impedance is likely to be similarly high for both topologies in the absence of feedback. The distortion cancellation in PP will only work on even-order products leaving the odd-order distortion (and possibly even increasing it). There is empirical evidence to show that odd-order is more subjectively objectionable, so perhaps not as beneficial as it initially seems. Feedback tends to decrease low-order distortion but will add higher order products at low level. Is this audible? Maybe, maybe not. Judging by the feedback vs non-feedback amps I've heard (SE AND PP) I believe it is.
It is possible to get some distortion cancellation in SE, between driver and output tubes - people have experiments with reversing the polarity of the signal from the driver to the output and measured a marked decrease in distortion connected one way over the other. This is however *extremely* unpredictable and I don't believe anyone who claims to be able to predict it. It's more luck than judgement.
I'm also of the opinion that JFETs can be rather good they're more linear than bi-polars - Nelson Pass has done a lot of work with these recently in his First Watt experiments, successfully using them without feedback. The F1 & F2 are interesting in that they are basically the same except one is SE, one is PP, both have pretty low distortion. The F1 being SE has a mainly even order distortion spectrum, while the F2 is PP and has a mainly odd-order spectrum. They have different flavours in the same way that PP and SE amps are reported to have.
I heard an F3 - which I thought a very fine amplifier (it user Lovoltech power JFETs). Switching to a 'high distortion' SE tube amp brought about a significant
subjective increase in clarity. Go figure. I would love to know why, myself.
In summary, SETs can give excellent results due to the good open-loop linearity and simplicity (which I see as a virtue). I'm not blind to the faults, high Zout, highish distortion, but sometimes the cure is worse than the disease. I'll use whatever gives me the best subjective results with music.
I'm not going to get into the other part.