I would be careful with trusting the answers to those questions.
People remember things to be more positive than they were. This is an established bias in human psychology
Fair point; to correct for this bias, ask n people to rate their experience with software reliability, then wait five years and do it again. The larger your n, the more you correct for individual variation (via the Law of Large Numbers)
You're right. Better keep to not asking and just weirdly spy on them, concluding the emergency recovery buttons are never used, and remove them.