It's not an improbable claim so much as a nigh-unfalsifiable claim.
I mean, imagine the following conversation between two hypothetical people, arbitrarily labelled RZ and EN here:
EN: By finding enough "code words" you can make any criticism of Obama racist.
RZ: What about this criticism?
EN: By declaring "epic", "confirmation mess", and "death blow" to be racist "code words", you can make that criticism racist.
RZ: But "epic", "confirmation mess", and "death blow" aren't racist code words!
EN: Right. Neither is "food stamps".
Of course, one way forward from this point is to taboo "code word" -- for example, to predict that an IAT would find stronger associations between "food stamps" and black people than between "epic" and black people, but would not find stronger associations between "food stamps" and white people than between "epic" and white people.
I think "nigh-unfalsifiable" is unfair in general when it comes to the use of code words, but I'm not familiar with the facts of the particular case under discussion.
Here's the new thread for posting quotes, with the usual rules: