Interesting, is there a reason why Postgres wouldn't just apply the shared query planning to INSERT...VALUES as well, it seems like almost a no-brainer to speed up all queries. Or is this more related to how the parser works with the values?