Why do some cresc.s/dim.s stretch to the next measure and some don't?

Impossible to tell from a picture, score needed

score please...

It's a conscious choice based on discussion here and advice from Elaine Gold in "Behind Bars". The engraving standard says that normally hairpins ending on the first note of a measure should not cross the barline. But if the first note "unusually far" (words to that effect, I don't have my copy of Gould handy to check) from the barline, then it is useful to extend it. The example she shows involves a key signature pushing the first note over. So we implemented it to check the distance to the dynamic on the first note and extend it if it is more than some distance, where we picked the distance empirically (by testing different values until it felt right in a variety of cases).

In your case, the stem direction isn't the determining factor in itself, but apparently it does have a subtle effect on the position of the dynamic marking below. Probably no more than a stem width's difference, but between that, the accidental, and whatever is going on with other staves here that is pushing everything further right than it might otherwise be, I guess it was just enough.

So, you could consider just moving the dynamics slightly left or right - that's probably both easier and safer than editing the length of the hairpins.