IIRC my N settled at around N=2.7, which means that a lucky first click would abort the queued random clicks between clicks two and three.
After that, the most important decisions for a fast match were deciding which mines not to tag because the final auto-tag would get them anyway and which empty fields not to uncover because they would be uncovered faster with an "uncover all surrounding" click on a saturated number. Clicking was definitely the bottleneck for me, so the main goal of the eye/pattern match was to keep the click-queue filled at all times. The eyes would always focus somewhere close but not too close to where the clicking happened, so that the pattern match could work on a stable image. Or sometimes just rest for a few clicks, when the hand could not keep up.
Fond memories of stupid things.