Most importantly of all: polls in 2016 were not weighted by education. Everyone in the polling world now knows that was the big indicator as to why polls were 'off'. (Again, even if they were 'off', the were still technically within the margin of error at the end, so it's not like it was too bad.) Now they know, and account for it.
A big example of pollsters correcting their polling misses was seen in the UK last year. After 'missing' the polling on Brexit and Teresa May's election, they course corrected and fixed their models. The result was in 2019 it showed that the Conservatives would win a substantial majority. Polls all throughout showed that. And despite what many in UK Era wanted to believe, Corbyn was uniquely unpopular and polls showing a tightening of the race were way off. Polling done long before hand proved true: Conservative won massively that night. Polling caught it, because they adjusted and learned from their mistakes in the past. We're seeing the same here Stateside.
I've said it before, but I look back to that Virginia election in 2017 as either the canary in the coal mine for what was to come. Despite being a local election, it had the nation's attention and as such was nationalized. Why? Because the whole thing became a referendum on Trump and Trumpism, even if no one stated it outright. Though everyone treated it that way.
That Governors race? Many polls had Moderate Democrat Northam up on Trump-lite Gillispi(sp?) by like 10 for much of the race. The polls tightened as time went on, and a few polls had the thing as a toss up. The end result? Northam by like 10. It was a complete rout. More importantly, it was seen as a rejection of Trump.
I'd say 2016 was a referendum on Hillary Clinton. 2020 is going to be a referendum on Donald Trump.
I'm tired; thank you for coming to my TED Talk.