Wisconsin's numbers are the most surprising to me.
Someone noted that the dam really broke in NJ for Republicans. This isn't that surprising. Their Republicans rested on the moderate Republican base in the state, which is a lot smaller nowadays and I think the Trump "tax cut" which really hurt UMC people in blue states like New Jersey put Dems over the top there. I bet if not for that tax cut, Republicans would have held on to a couple districts there.
I think these numbers are the most telling in Pennsylvania because they had a fairish map and fewer incumbents. But Dems also had good candidates like Conor Lamb who did better than the Dem nominee will likely do in that district.
Although the tax bill was certainly one factor that contributed to certain districts flipping, I would hardly consider it to be the predominant issue that the seats you mentioned hinged on (healthcare + Trump's personal unpopularity with that demographic were probably far more decisive).
I think if
either the tax bill
or the attempted Obamacare repeal isn't a factor, MacArthur holds on; if
both aren't factors, Lance holds on; if both aren't factors
and Trump is either not President or not personally repulsive to the upper middle class, Frelinghuysen doesn't retire and either beats Sherrill by a hair or loses by a hair (depending on whether it's still a midterm with an unpopular Republican President or not). The dynamics of the race to succeed LoBiondo were such that I don't feel comfortable speculating on counterfactuals for it.