Frankly, this has an obvious answer: The people who answer polls are not representative of the population at large. Yet this was known decently enough since 2016. But more at the heart of the issue here is pollsters are providing a service not to inform but to persuade (both the public and donors); internal-polling is actually what matters. And instead of reformulating the service with more accurate questions for total populace (who will your neighbor vote for?), the pollsters doubled down on the standard methods and made (roughly doubly!) a larger polling error than 2016. I mean the Maine senate race was on average ~15 points off the final result[0]. Frankly this was the race for the pollsters to prove themselves that they can self-correct, as they have demonstrated to be unable to do this, it is clear they just should be abolished.
What I mean to say is polling as a major industry for political campaigns is done after this election. The current people running these public polls have no place of confidence anymore: donors should not take them of value, and public opinion should ignore polling. They had their chance this time around for survival as an industry but roundly threw that away.
[0] https://www.realclearpolitics.com/epolls/2020/senate/me/main...