The issue one could reasonably take with the Arizona poll is why so many more people id'd themselves as Demos than Repubs, which doesn't reflect the known party affiliation. One could argue that Repubs were more likely to hang up and not take part than the Demos, and out of those who did, most or all of them would vote for Trump. We don't know if that would have changed the Repub numbers up in favor of Trump.

The NBC poll you alluded to above (D+11) also has the same issue with sampling, as well as the issue of being run by pro-Demo people. So ya, that poll should be treated with suspicion.