Revisiting previous comments:
The EAT 'cool' valves: ........
Again (as always?

) I am sceptical. Nothing I could find shows a definite advantage
for audio.
[My favorite phrase, but to be fair it must be acknowledged that 'better' performance under certain extremes often (mostly?) translates to nil for audio. E.g. some shock tests; unless you are going to use your amplifier in an army tank barreling over rocks and through ditches those are irrelevant under home conditions.]
Thus to pay that much more for practically mainly somewhat improved heat radiation; sorry, count me out. Not to re-start another point of contention, but similar hype reigned regarding cryogenic treatment some time ago; died a quiet and welcome death since.
In electronics, to make a rather sweeping statement: The pattern is always roughly the same: A grand term suddenly pops up or is regurgitated, its often imaginary advantages are lauded by the manufacturers (how else), and a small group tags along nodding their heads up and down, confirmation by the wife-in-the-kitchen, the gardener, the family dog (apology to F_D!) ....... To stop there.
Not that improvements cannot still be made particularly with new materials, but somewhat sterner proof is needed than "I tied a knot in my speaker cable and it sounded better."
Explanatory:
The working of normal valve testers need also be considered. To my knowledge they mostly use 'pulsed' voltages as in half-wave rectified ones. The steady reading on a meter is then really an integrated "average" of a half wave sweep. Because valves are non-linear this will give different Gm readings than datasheet stats, which are mostly taken at steady-state values of the listed voltage or current (pulsed in case of high values to accomodate dissipation limitations). That does not imply that a valve tester is useless; given parameter values are often adapted to give realistic results of the valve's condition. But that data cannot be used for design purposes.
Coming closer to Mr Lafferty's and Dr Gillespie's tests: The reading of distortion at high levels is valuable, but a mid-value measurement is also required. Small differences in valve characteristics will cause overload to occur at different points, thus seemingly significant differences could be measured at maximum output, one valve being just shy of overload while the next one is just in it. As Dr Gillespie mentioned, the sample was too small for measured differences to be tacked onto the brand itself (see the differences found between EH valves under different conditions).
Complex? Not really, if one just understands what exactly is being measured and what not.