This miss on El-Sayed's polling was also pretty severe but since it was still on the correct side of the outcome it hasn't gotten as much attention as this will.
Polling in the US is collapsing. I thought it would have happened before now, but it's finally getting there. Telephony in the US has changed so fundamentally and completely that there is no putting that toothpaste back in the tube. You cannot get a good poll without a good sample, and people will not allow themselves to be sampled via phone anymore because the old 'answer the phone!' imperative is totally gone today. That was the thing that made polling work in the pre-cell/pre-sms era - that all people would answer their phone. When people no longer feel that panic about letting a ringing phone go unanswered, the paradigm shifts from a random sample to a self-selected one - and you are hosed as a pollster. There is no current technology or method that will get you a non-self-selected sample in the US anymore.
So what we are left with is pollsters making adjustment to polls that drive their reported result more than the data actually in the poll. So if a pollster is a good prognosticator, he gets the adjustments right and has a good poll. But who can predict from cycle to cycle what the right adjustments are? That's a rear view mirror process and by nature is bound the miss the very thing you want to know in a political poll, which is "what is changing".