Morning consult is an ok pollster at best. It was graded B by Silver if chatgpt is correct, and is currently ranked number 116 in Silverless fivethirtyeight this year:
https://projects.fivethirtyeight.com/pollster-ratings/
Big Village is crap (Grade C)
The rest of the polls you posted are not reassuring, rather worrying. Harris being 2 or 3 ahead nationally is not good enough because of, again, the Electoral college bias against the Dems. Biden was ahead by 7 and 8. It is not 2020 redux at all. The race wasn't that tight back then.
Harris still has a chance of course, a tight race can go either way, but Trump beat his polls twice making up for the gap with Clinton at state-level, and losing several states to Biden at a margin way narrower than predicted by the polls. If this repeats itself Harris is doomed, if not it's currently around 50/50.
I can quote another 10 polls in the last 2 weeks, which would be redundant. Anyway, from my point of view, what has happened since the convention, polling wise, is:
1) A deluge of known, low quality right wing polls aimed at flooding the zone and turn the narrative
2) A dearth of high quality state polls (understandable, those are expensive to do)
3) Static or a slight uptick in Trump’s national number due to RFK’s team up
4) A very wild swing of number even between renowned pollsters due to their sample/weighting. ABC/Ipsos 2 weeks ago have Harris +6 with LV with a D+6 sample, NYT/Siena have her down one with R +3. The average, however, is pretty steady high 3s-low 4s.
As for the rest of your concern, I don’t subscribe to the idea that just because polling overestimated D vote share twice against Trump that it means they will do the same this cycle, and we should just automatically subtract a few points from Harris as a safety measure. Because a) that’s not how stats work and b) we have plenty of evidence to the contrary that polling in general have swung the other way since Dobbs and overestimate Republicans, even Trump underperformed his own numbers in the Republican primaries. Every other metrics, from predictor like the Washington primary result (
https://www.newyorker.com/news/q-and-a/how-kamala-harriss-coalition-changes-the-race-for-congress), fundraising number, ad buys, campaign activities point to a tight race with the Trump campaign zeroing in a PA+NC+GA narrow path at exactly 270 EV. It’s certainly not the picture of a tied national number if we do the subtraction exercise because in that case you would see them campaigning in Minnesota and Virginia, not withdrawing from NH. It is 2020 redux on the actual ground, even back then Democratic pollsters and surrogates have repeatedly cautioned against rosy public polling, Bernie Sanders went on TV to warn about election night early Trump lead due to Democratic mail in votes, it’s all on the record.
And I agree, D+3/4 is not safe, Biden at 4.5 would have lost with 7 million votes margin if 42000 swing states voters voted the other way, but it’s the reality we live in. Obama’s blowout in 2008 was +7, in a much less divided electorate and every headwinds going against Republicans, you simply won’t do better than +4/5 at best nowadays. The electorate is so baked in that we end up praying that a few million voters in a bunch of states vote the way we want every 4 years, but it would be the state of play for any candidate in a competitive race regardless due to hyper partisanship, so fretting about it from an imprecise tool as polling is a pointless exercise, much less off a single poll. Going forward, the only way to avoid this is the Nat. Vote Interstate Compact, because a Constitutional Amendment isn’t happening and demographic change in the Midwest means 2016 will become the default more often than not.