If you read the linked article at the bottom of this article, you find out that this number only holds up for Postgres. It’s a cool hack, but this is all just working around deficiencies in the query planners in various databases and so its effectiveness will depend on the planner in question.