The idea of R is that it allows you to reason about things like "I need more than X input precision to achieve Y bound on my output". Just sticking with naïve

The idea of R is that it allows you to reason about things like "I need more than X input precision to achieve Y bound on my output". Just sticking with naïve computation in Q does not suffice for that.