Texas survey was Trump +8 sample.
Strong evidence here that PPP is either cooking their books on that recalled vote statistic or are getting a wildly unrepresentative sample that massively overstates the Trump/Biden switcher. The real electorate of Texas swung 3 points Democratic this election, but according to PPP it was going to be a 10 point swing. I get recalled vote is not super reliable, but I don't know how in the hell anyone can make sense of that.
I think that most polls were way oversampling Trump-Biden voters and that was a major source of the polling error. There is no way to solve it by weighting. A pollster can be as thorough as possible, but if for every possible demographic characteristic a person is disproportionately likely to vote for Biden than reality then there is nothing to be done. Pollsters were probably thinking, well we got the right number of Republicans, right number of non-college educated voters so we're good.