To start, I'll define scientific realism (hereon called "SR"). As I mean it here, SR consists of the following set of propositions:
For example, that there really is such a thing in the physical world as the strong nuclear force, that it is carried by gluons, that those gluons have color charge, etc.
With that definition out of the way, consider the following: Imagine an alien civilization with radically different linguistics to our own. All concepts and thoughts can be expressed by drawing single glyphs. Anything from "Hi," to the entire content of Das Kapital, to the emotions of the scribe, and indeed to entire physical theories are expressed in a single character. The alien glyphs are, importantly, not just a nonlinear writing system. There are no patterns to the ways they are composed. No human could hope to ever learn such a system beyond memorization of very basic phrases. The way to interpret the meanings of the symbols, however, and how to compose new ones, are understood by all of these aliens from birth onward.
The physical theories these aliens have created have the exact same predictive power as our own (as well as having whatever other theoretical virtues one might propose). Given that, it would seem theirs have just as much claim to the truth.
Now, to the core of my argument. One of the following propositions must be true:
In light of (1), (A) looks pretty undesirable. Alien physics and human physics are irreconcilable. They have entirely different forms and fundamentally contradictory ontological commitments. Indeed it's not even clear what it would mean to reconcile them. If alien physics is even approximately true, then human physics is massively in error, and vice versa. To accept both (A) and (1) then, it seems one is committed to believing that there are true contradictions, as the contradictory ontologies of alien and human physics would be equally true. Most realists, and indeed most philosophers in general, are not willing to say this of course.
(B) obviously doesn't look much better. I hope I've demonstrated that the person who accepts (B) would have to make a completely arbitrary choice as to what theory to endorse given that human and alien physics are predictively identical. The only way to maintain SR would be to arbitrarily say that only human physics is right.
(C) and (D) blatantly contradict (2), and therefore would entail the falsity of SR.
So then, SR is either contradictory, arbitrary, or false. Personally, I think (D) is most convincing. In the end, I hope I've provided a good case that there is no both non-arbitrary and non-contradictory way to be a realist. Though honestly, as if the world needed another reason to dismiss scientific realism. :o)