Criticizing subsamples doesn't necessarily make any sense.
I can't recall the poll board ever being this bad. Everyone just pretends as if every poll going against their candidate is useless with arguments of ever-decreasing quality.
I agree with you when it comes to party ID etc, but I think there is some credence in being concerned with demographic sub samples which don't change nationally by too much year on year. I mean this poll is pretty consistent (when compared with other polls) when it comes to White and Black voting intention; but it sticks out like a sore thumb with Hispanics. It would be the same if a poll suddenly showed Romney within 10% of Obama on the black vote despite that not happening in any national election in the modern era.