Objective reality, we might think, is evidence-transcendent. It goes beyond simply how things appear, to us, to be. No matter how careful our inquiry, the evidence may be misleading. So our justified conclusions may still diverge from the truth of the matter. This is in tension with any view that defines truth (perhaps in a restricted domain of discourse) merely as the end-point to which rational agents would ultimately converge. We might reasonably expect to converge on the truth, of course, but we shouldn't rule out from the start the possibility of insuperable error. Not if we take ourselves to be describing facts that hold independently of anyone's opinions about the matter.
Is that right? Or could there be a domain of objective reality that is transparent to rational inquirers, and this fact of transparency is itself transparent? For in the latter case, we could be in a position to know that ideal judgments couldn't diverge from the truth after all. But that needn't lead us to deflate its objectivity at all. The order of explanation still comports with realism: ideal agents would agree on P, presumably because P is (independently) true. But perhaps in this case the preclusion of ideal error is not really made 'from the start'. It's only after judging that we (somehow) have epistemic access to the independent metaphysical reality that realists can make such claims.
In contrast, quasi-realists (idealists? constructivists? I'm not sure what the best label is) reverse the order of explanation, and hold that P is true merely because ideal agents would converge on this result. So the notion of ideal error really can be dismissed right away, simply by definition, before we begin any kind of extra-linguistic investigation.
The modal rationalist -- who links possibility with ideal conceivability -- must decide which of these two paths to take. The constructivist line seems almost eliminativist, though. Intuitively, I would think, our concept of possibility is realist in nature. We think that possibility claims are objectively about the world (and how it could have been), rather than merely about our idealized opinions. But if we want to be realists, do we have any grounds to rule out the possibility of modal error persisting to the limit of ideality? Or should we instead be skeptical of whether ideal conceivability suffices to guarantee possibility?