It was around the same for me: I knew I had to be careful for the first problem (that was accentuated by the fact that we were warned about the failure rate). For the second one though, I was too careful: I immediately started transcribing the problem in a pertinent format, i.e. I created the machine.minute unit (equivalent to the kWh unit) that allowed me to understand that a widget is made in 5 machine.minutes. Then I looked at the question, and computed that you’d need 500 machine.minutes to make 100  widgets, so 5 minutes with 100 machines. During that last step, something was bugging me, and it was the realization that the problem was much simpler and intuitive. I still finished what I was doing, then looked at it as a whole, and saw the "you need more widgets but have more machines so you need the same time" intuitively.
And for the last one, it was very obvious of course.