It's like if you were trying to evaluate the average math proficiency of the population by only reading blog posts about math. It would seem like everyone really knew math! Except of course, the only people who blog about math are the few that are into it. They're not bullshitting, they're just overrepresented.
Not that I disagree with your original point, by the way; BS is rampant.
I agree with this. I have interviewed a good amount of engineers, both as a SWE at Google as well as founder of my own company. From that experience I can tell that the only thing that asking mostly hard interview questions does is reducing your candidate pool to 0.
Even with mostly moderately hard questions your candidate pool at the end of the funnel is probably going to be around 5% to 10% as compared to the start of the funnel. Plus after a certain level of question difficulty as an employer you're probably optimizing for the wrong skillset.
Perhaps for super competitive positions, e.g. for algorithmic trading at reputable Wall Street hedge funds, there's a place for killer interview questions but it's definitely not the bread and butter.
Spending 2 years learning how to interview when you could have been spending 1.96 years getting passably good at something and 2 weeks on interviewing is madness, though.
You can imagine how scary it is to read all the super tough questions, going to the whiteboard, having to discuss passion projects, needing a solid github profile, knowing algorithms, data structures, strange math, and on and on.
You start interviewing and you see that none of the above happens 99% of the time. You wasted 2 years learning a bunch of stuff when you really should have been interviewing six months after writing your first line of code.