I'm willing to give the leaker the benefit of the doubt, at least until some other body of evidence shows him to be wrong. Nobody has shown power numbers yet. If he has a leaky-as-hell ES (which would explain the low voltage) and a water setup using something like one or more MO-RA3s (or hell an old copper auto radiator) then yeah he could be dumping 400-500W of power from the chip while still keeping temps relatively low. It is possible.
Versus chilled water, for example. If he's using a chiller then it's a whole 'nother ballgame.
This. I believe the results, i would be surprised if they turned out fake. But its almost 100 percent thanks to several factors in play, like golden chip and some serious watercooling setup in action. Surely you wont get the same results trying the same with a retail chip at home running run of the mill 240mm AiO.
BTW, people seem to think, that if OCed 7900x was difficult to cool, 7980xe will be borderline impossible (without LN2). While it may sound logical, its probably very likely not to be true - and these results are the first glimpse to show that. I recall 5960x clocking pretty much as high as 4-core Haswell chips, so IMO its more about the underlying chip design/process for the current generation, rather than number of cores. Similarly, I could OC my 6850K to 4,2Ghz max (on acceptable voltages) - now i am not accomplished overclocker, so perhaps someone more skilled could get out of it more, but from i could gather, its IMO safe to say, the 6950x with 4 more cores did not on average clock that much worse overall.... This could be similar case.