Ben <bentenzha@xxxxxxxxxxx> <MWHPR06MB2400E1FA4AAD12440EA96B79B93E0@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> writes: > You are right none of them is what I originally used to create them > which should be > `select * from utlog.stats_per_shift_filtered_b0206`, but they do look > different, This discrepancy could explain a lot: > (stats_per_shift_filtered_u0206.wsft)::character varying AS wsft, vs > stats_per_shift_filtered_b0206.wsft, Your problem query has a WHERE clause testing wsft, so this difference is subtly affecting what that WHERE clause means. That could result in a plan change, thereby changing the row output order, which as already noted would be sufficient to explain the inconsistency. If I had to bet I'd suspect that the extra cast is blocking use of an index on wsft. regards, tom lane
