The 1000m. (if we are sticking to events that have IAAF/WA WR status).
If you look at the plotted graph it would also suggest that the 1000m is under-indexing vs the 800m and 1500m head to head as it is, but accounting for a normal curvature the "20 second barrier" would land in somewhere in the 1000-1200m zone.
The sub 100 800m just seems like a bit of a pipe dream right now, especially as we have almost exhausted the ways to execute the event for optimum performance. We had Kipketer running low 1.41 (.24) off an insane first lap (48.3), we had him run marginally faster off 49.1, Rudisha high 1.40 off a low 49, Wanyonyi in basically the same ballpark off 50.0. Maybe 1.40.91 get's beaten (I feel the WR should have been in the 1.40.6/7 range from Kipketer anyway), but 1.39.99 is 7.3 up the track from Rudishas mark. I don't think so.
1500m - we haven't had anyone, clean or not within even half a second of El G's second best time* of 3.26.14 in 24 years. That's despite a shift in the event philosophically which has improved performance plateau en mass, coupled with wavelight (creating a link between what guys specifically work on in training and being able to execute in a race), bicarb and better footwear that doesn't beat them up as much. It would be a miracle if someone ran under 3.26.00 - where is sub 3.25 coming from?
So the 1000 - run even more infrequently than the 3000, WR at 2.11.96. 1.97 seconds needed to get to 2.09.99. Could someone get out in 1.16.5 and close in just under 53.5? It just feels like if we ran as many 1000m races as 8's and 15's this record might be in the 2.10.5 ballpark already (1.44.0/26.5) and here is the interesting thing - a 2.10.5 1000m converts on the tables (not perfect but really quite good giving us comparisons) to a 1.41.40 - a time which Wanyonyi and Arop have both bettered and Sedjati is within 6/100ths of.
For me it's the 1000 - not even close, and it just seems like the perfectly "round" distance - an exact kilometer, to align with this challenge.