There's been a rule of thumb the last ten years or so. Republicans do very well in VBM voting before election day, less well in election day voting, and poor in ballots counted after election day. The theory is that Republicans tend to vote heavily with vote by mail. Those are the first ballots counted. Democrats, on the other hand, turn in their VBM ballots on the day of election and are more likely to be provisional voters. Those are people who might have a VBM ballot, but vote in person, or they aren't on the voter roll, yet claim to be a voter. I don't know the psychology of all this, but it's happened almost universally with each election since I started following California elections in 2010. So a Republican wants to be at 57% after the first ballots are counted and at 52% when election night ends. If neither of these happen, the Republican is likely to lose.
This year's California primary has been different. The Democratic pre-election day VBM were stronger than usual and the NPP/other part ballots were stronger than that. I've been tracking 40 elections (17 congressional, 7 senate, 16 assembly) that were either expected to be competitive or were a little out of that range. I excluded districts where there wasn't a Republican or Democrat or ones that weren't expected to be competitive. To see which districts might be competitive in November, look at the same districts in June. The spreadsheet is here. In 31 of the 40 elections Republicans did better on election day than they did in VBM ballots. Post-election day counting isn't over yet and it may still change. It's possible that some counties in a district that might be more Republican have counted faster than those that might be Democratic. Republicans have gained in 25 of the 40 elections, however. In the past they might've gained in 4. So we've seemed a voting pattern change. This tells us that some of the automatic assumptions we've made in the past could change.