1. Many segments have duplicates. I think a fair assessment of this is, if the top X riders are the same, the segment covers the same route and the stats (elevation gain etc) are the same (or very similar to 1%); then the segment is a duplicate. How many riders is X? I'd say the top 10 or top 50. Anyone outside the top 50 I can't imagine being too upset by a minor reshuffle based on a slightly different segment. Older segments tend to be constructed using less accurate GPS and many riders have been over the segment with much more accurate GPS data, most prominently, elevation. Too many times I have tried for a KOM on a segment thinking it's got a false flat, levels out or kicks up only to find the strava topography was very off. When I look at my GPS, it shows a topography different and much more accurate than the one associated to the segment (presumably the GPS of whoever originally made the segment). It would be good if Strava could update the segment GPS so it is accurate to the topography. I put my phone on my back, which means a very clear signal.
2. Strava could probably reduce the minimum length of some segments by using GPS points instead of distance. The message "segment too short" could be replaced with "segment doesn't have enough GPS points" when creating a segment. I think this would be better for slower speed segments like running segments or MTB segments. Strava could probably use the KOM time as an estimate to know if the segment is actually too short. My suggestion is that any segment with a KOM time under 20 seconds is "too short" regardless of it's actual distance. More often than not, with segments with a 20 second or under KOM the top ten has at least 5 riders on 2nd place with the same second time, e.g. 1st 20s, 2nd 21s (for 5 more riders) and the 10th place is like 23 seconds. It's not a realistically useful leader board.