First off: apologies if this belongs elsewhere. I found a Strava article about how to post ideas, but it seems out of date, because I can't find any of the pages/buttons mentioned.
With that out of the way: I recently saw some segments where the record holder is listed as completing the segment at average pace X min/km. Then, looking at the details of their segment attempt, it was clear that their actual pace was way lower (Y min/km).
Upon looking at their recorded track, it was clear they turned around after 70-80% of the segment. It seems Strava matched their partial segment to the complete one, and then calculated their average pace as “their time on this segment” / “actual segment distance". It was obvious that the pace “their time on this segment” / “distance of their recording on this segment” was very different.
My suggestion is to use this latter as a filter for mismatched segments. Clearly the athlete didn't mean to cheat here, but the Strava matching algorithm misattributed the KOM to them. By comparing the two measures of pace mentioned above, it should be fairly trivial to filter out gross outliers (e.g. actual pace < 0.8 * “calculated segment pace").
