Who knows?
Seems doubtful. According to Dave Leip's US Election Atlas results page it hasn't happened since 1948 (I saw that you already wrote that, but I checked it just to be sure). There seems to be no particular trend when comparing the results from 1996 to 2008. (1992 is weird because of Perot's strong showing.) From 2008 to 2012 there was a big difference. Obama won PA by 10 points in 2008 but only by 4 points in 2012, but in both cases the votes for Obama were higher in PA than nationally, and both time by around 1%, so there was not trend in PA that wasn't also the case nationally. Also, over the three-month period that we have known who the two major-party nominees are, the polls have ranged from Clinton +0.5% to Clinton 8.9% for PA, with the exception of one Quinnipiac poll from early July has indicated a Trump plurality of 2% over Clinton. The latest I saw was Clinton 45% to Trump 36% (NBC/WSJ/Marist July 11). During that same three-month period, the range was Trump +.2% to Clinton 6.8%, and a CBS poll published around July 11 also showed Clinton +4%. I have not seen any polls of PA since the GOP convention ended. It would be interesting to see if the national Trump bounce was the same as, greater than, or less than the Trump bounce in PA.
Hard to say, but there seems to be no reason to think that the Trump percentage in PA will be greater than the Trump percentage nationally, if that's what you're asking. Much of it depends upon turnout if the talking heads are to be believed. Suburban Philadelphia and inner-city Philadelphia votes for Clinton and much of the rest of the state votes for Trump. All the major outlets are calling PA a "swing state" now. Anything can happen, I suppose, and there's really only one poll that matters. We'll have to wait till November to answer that one.
The major thing you're missing here is that if Latinos swing hard to the left (a big if, obviously), that will probably be enough to move PA's PVI to the right by default. And if it was only a 1% D PVI before, it wouldn't take much to push that to R.