No, that isn't how it works. You clearly aren't understanding "basic probability theory".
No he isn't. More specifically, the 'theory part. IE: Just because the theory states that, with a large sample size you'll find the probability to be correct, it doesn't mean it always is.
I do wonder what the % it's come out with overall is for the population of single server, though, as that would certainly be classed as a huge sample size.