Polling outliers.

For real. How hard is it to say "this poll appears to be an outlier from the others" or to show the margin of error for a the subgroup in question? Or to not report on a crosstab at all if the sample size is too small?

For a presidential poll, the reported margin of error only applies to the topline of the poll- the overall poll. And the candidate vote share at that, not the race difference.

For a subgroup, say black voters or young voters, the margin of error will be larger as the sample size is smaller.

And the subgroup may not be weighted like the overall poll. So you may get really weird subgroups in polls where the topline is perfectly normal.