matthewr
spɹɐʍʞɔɐq spɹoɔǝɹ ɹnoʎ sʎɐld
The whole country diverged from the 2017 results, then converged on them again. What's more likely: that the results this time resemble the 2017 election, or that they resemble the 2019 European election?
That's a bit of a straw man though, I think, as nobody outside wishful thinking Lib Dems takes too much from the 2019 Euro election and the various models are based on 2017 (over estimated Tory vote), 2015 (under estimated Tory vote) and previous general elections. There are also more things you can look at from 2017 that tell you something about whether this time is likely to be the same. E.g. Labour benefited from a significant rise in Corbyn's ratings and a collapse in Theresa May's which seems absent this time despite the awfulness of Johnson. Focusing on 2017 looks, to me, like a way to convince oneself that things will be good for Labour.
Although I do agree with your later point that there is so much uncertainty here that all a voter can really do is look at local conditions and campaigning and be guided by that.
BTW A bit harsh on the Observer about their article on marginals I think as that is Peter Kellner writing so it's more like his and YouGov's expert view than that of Guardian journalists.