I have interviewed a lot of people. Most of the time I can determine three questions in if someone is not a good fit. This is on a technical interview. I have never done an interview where someone blew, or gave middling answers to the first three questions, and then recovered. Also, I never gave an interview where the person hit the first three questions out of the park and then started flagging, in terms of technical answers any how.
In terms of people who did well on the first three questions but who did not get an offer, it was always due to "cultural" issues, and it was someone else (often the manager) who raised those cultural issues, with me sometimes disagreeing (for instance, my old manager thought someone was too arrogant, whereas I didn't mind).
They're ones I've asked before.
They are questions that have gotten some coherent response from about five out of six people, but generally only get a good, clear, robust response from one out of six people.
Usually they touch on different areas - I don't ask three questions about function parameters, I ask about very different topics, as theoretically they may be weak on one area and strong on another (although usually people have the same level of strength over all areas in my experience).
I don't really worry about asking the same questions to each candidate, since if they hit the ones I always ask out of the park but then struggle on ones I only sometimes, or rarely, or never ask, I would find that odd.
I might adopt questions to what they have on their resume, but generally it takes a while to adapt to a newer question. If we start using a new language more and an old language less, I probably will adapt questions to the new language and so on.
i like the approach, do you just outright ding someone due to the googling of answers offscreen?