Overall the polls missed in 2020 by about 3.5% to the blue side. If Biden were to win another elections by exactly the same vote it would mean that error had reversed to a few percentage overestimate to the red side. So if you believe that US politics is actually pretty ossified, one explanation would be that the pollsters looked at the miss in 2020 and when they re-tuned their likely voter models they went too far the other way. But even so, 3% is a pretty standard margin of error of political polling, and in the US electoral system 3% of the popular vote easily swings an election from big loss to big win. Add to that the big turnout uncertainties you're at a point where too much fixation on the polling is sort of futile. There is enough uncertainty for the election to be out there for either side to take despite the polling data esp because no-one has a good handle on who is going to vote anymore.
Polls present risk either way. If you trust polls that show you ahead but are wrong, you don't work hard enough were you need to, if you trust polls that show you behind that are wrong, you pull resources out of winnable districts. Just run the best campaign you can.