For example: UPDATE table1 SET status='expired' WHERE id in (SELECT wrong_id IN table2)
This will update every row in table1if wrong_id doesn't exist, ignoring the ERROR: column "wrong_id" does not exist from the subquery.
The subquery never provokes that error by virtue of the fact it is a subquery. It's only if you run that as a standalone query do you see the error. This is because correlated subqueries are a thing (and, yes, they are documented).