We have two RDS environments with read replicas (test has one, production has three) and a conundrum: why does the same query, over the same data + same indexes, take close
