I don't know if you know this, but you can actually look at the polls and the sampling.
538's latest polls is really good because you can see a poll and click on it and see the sampling: https://projects.fivethirtyeight.com/polls/
I clicked like 10 of the latest and didn't see any sampling that you suggested.
Here's a recent MN one from Suffolk University.
https://www.suffolk.edu/-/media/suffolk/documents/academics/research-at-suffolk/suprc/polls/other-states/2020/9_25_2020_marginals_pdftxt.pdf?la=en&hash=EA09FFF4A22FBF2A04A6A5776C86103A990C9C6A
Has Biden 47, Trump 40.
With rounding it was 40 Dem/36 Republican.
Click any of the others, there's not any I saw with a 35/25 split.
A couple things on polls:
1. People say polls were way off in 2016. If you look state by state, most ended up being close and within the margin of error.
2. Look at RCPs averages in 2016, there were huge fluctuations between Trump and Hilary during the campaign. In 2020, everything has been pretty steady for Bidens lead. There really have not been a lot of fluctuations, especially compared to 2016 and how much that changed day to day.
3. To points 1 and 2, I think during 2016, we saw almost all the undecideds go for Trump (I think the actual split was like 90/10 Trump). The issue with this election is there are not many undecides. Someone posted a couple pages ago about how a survey showed Florida undecides were going heavily for Trump. When you pull the actual survey, undecides made up 3 percent of the population. I can see that being true as it seems hard to think there is the level of undecides. I mean really, who doesn't have one opinion or another of Trump by now? In 2016, we saw far more undecides at this point.
4. Polls were pretty accurate on the 2018 blue wave.