We just swapped from Query 3 (using le) to Query 4 (using not gt) and saw an immediate 10x (11.6x to be specific) improvement. Waiting on longer metrics to confirm it remains a 10x improvement.
I’m still curious if dropping the index entirely would speed this up.
Also, I’m curious if precision date equality/math is required, if the better/best approach approach is using an epoch / int field if speed is the optimizing factor. And if so, how much faster is it.