In 2016, polling was fatally flawed especially in the US general presidential election. One mistake was in sample selection.
At least one 2020 DNC primary election poll has had statistically insignificant samples to represent those under 50, which makes me wonder how much has really changed since 2016.
I would be interested to find metadata on primary and general election polls in 2016 and primary election polls in 2020. Specifically metadata on sampling selections and other polling features that may indicate if and how polling methods have changed or remain flawed.
Data for the expected 2016 and 2020 voter demographic turnout would also be useful to see how representative the polling sample selections are.