On prod, cold: Limit (cost=241.77..242.26 rows=28 width=76) (actual time=14.100..14.162 rows=4 loops=1) and then 20:01 < stub> 7ms on prod_2 20:02 < stub> 70ms on the master 20:02 < stub> 1.4ms the second run
So, this looks win to me.
On prod, cold: Limit (cost=241. 77..242. 26 rows=28 width=76) (actual time=14.100..14.162 rows=4 loops=1)
and then
20:01 < stub> 7ms on prod_2
20:02 < stub> 70ms on the master
20:02 < stub> 1.4ms the second run
So, this looks win to me.