So, I am open-minded but not super persuaded by this. There are a handful of counties to have counted provisional ballots so far and those ballots indeed went for Trump, but they came from counties where the rest of the vote was *even stronger* for Trump.
https://www.electionreturns.pa.gov/General/CountyBreakDownResults?officeId=1&districtId=1&ElectionID=undefined&ElectionType=undefined&IsActive=undefined https://twitter.com/Wizard_Predicts/status/1324809471068131333

Maybe the provisionals—which should mostly reflect cases where the voter originally requested a mail ballot, then decided to vote in person—are slightly D leaning but not as D leaning as mail votes. Maybe they're about neutral. They'd have to be strongly R-leaning to help Trump.

Also, about 40K of the provisionals are in Philly County, which is a considerably higher proportion than Philly's share of the vote overall. https://twitter.com/DeItaone/status/1324770591631704066?s=20

Furthermore, the universe of requested-but-unreturned mail ballots, which is where most of these provisionals originated from, was D +17 by party registration. Not as D as completed mail ballots, but obviously still quite D.

https://electproject.github.io/Early-Vote-2020G/PA.html

One other point of context here is that there *was* some confusion about provisional ballots earlier because York County incorrectly listed some of its provisional ballots as mail ballots. Those were very D, but they weren't actually provisional ballots.

Nonetheless, we have 3 points of evidence that the provisionals are somewhat D (party registration on unreturned mail ballots, concentration in Philly, comparison to other ballots in counties to have counted them) when they'd have to be *MASSIVELY* GOP-leaning to save Trump.

Another thread from local reporters and experts that comes to the same conclusion: https://twitter.com/julianrouth/status/1324822379072245763?s=19