> But wouldn't a plateu in performance also cause a plateu in record setting?
No, not if there's variability in performance and an expanding population. Records are set by the most extreme person, not by the 'average performance'. If you have a billion runners, even if the average runner is slower than before (performance not just plateauing, but diving), you may still wind up with some freak with perfect genetics and simian arms and obsessive personality setting a new record, because when you look at a billion runners, odds are one of them will be bizarre in just the right way to set a new record.
An aside, thanks for the EVT reference. Something new to learn.