"All"? Testing variations among samples of speakers is not the same thing as testing in stereo. It would double the work, at a minimum for two samples, and isn't really the point at this time AFAIK. Just testing one is quite a chore. I'd rather the time be spent testing more, different, speaker models than attempting to asses manufacturing variability.
Yes, understood, of course, and I agree, building up a bigger database would be good, albeit with certain data unreported. But I didn't ask for anything to be tested in stereo - I said good stereo performance includes but is not completely defined by attributes apparent in mono, so let's expand the mono testing to include a couple of extra headline items, to give us some extra guidance.
The argument against doing that is about cost, time, and inconvenience, not about engineering - always the limitation faced in a hobby context, as opposed to a professional context. Can't we admit that, instead of trying to pretend it's something else?