I think that Trump being on top of the ticket will probably increase the magnitude of any polling error, since it appears that one major issue with 2020 polls was that Trump supporters were significantly less likely to respond to pollsters. Nonetheless, polls have had greater misses in the past, and if 538's polling model were applied retroactively to past election cycles dating back to 1972 (the earliest year with sufficient public polls for that model), significant polling errors (4 points or more) occurred in 1980, 1996, 2000, and 2012, in addition to 2020.
This is interesting -- do you have a link?
