Right now we check that the outlives requirements in the trait are met, but say nothing about an associated type. I think the correct thing is to treat them exactly like type parameters, basically. So we should be checking that they outlive the relevant region bound -- and we should assume that they outlive the current fn, but nothing else. This seems to be roughly equivalent to the current treatment of type parameters (for better/worse).
Right now we check that the outlives requirements in the trait are met, but say nothing about an associated type. I think the correct thing is to treat them exactly like type parameters, basically. So we should be checking that they outlive the relevant region bound -- and we should assume that they outlive the current fn, but nothing else. This seems to be roughly equivalent to the current treatment of type parameters (for better/worse).