Hum... If you know `a` is a key of table `X` and `b` is a key of table `Y`, you can replace a `select distinct` with a `join ... on exists ()` with complete certainty.

And this solves almost all of the problematic cases of this on practice. If the thing you are querying isn't a key, you will naturally think a bit further and not write the distinct.

I guess that kind of optimization just goes counter to the philosophy of SQL optimizers. I know that I would be deeply surprised if I found a database doing it, and not in a good way.