"Polling was conducted using a clustered IVR approach and results were
weighted to reflect the perceived electorate based on past presidential
election trends in Arizona. "
Why should we expect the electorate to mirror "past presidential election trends" in Arizona? I think polls that rely upon "voter files" are usually the most reliable but there are instances in which they have been miserable failures. In the event of a surge in voter registration and concomitantly high voter turnout, a reliance upon a voter file to create a universe of "likely voters" will almost certainly fail because irregular/new voters will compose a large share of the electorate.
How I know this poll is garbage: Donald Trump's favorables are far too high.
I don't know. Hope you're right, but that's precisely the scenario that should keep us up at night.
As it turns out, I was wrong. I forgot to check the national polling aggregators; Trump's favorables have surged over the past weak from -25 to -15. Therefore, the poll probably makes a degree of sense.