If someone's targeting me specifically then I feel like it's easier to just watch my house in person. Plus there are all the neighbors with (I assume) their doorbell cameras that would also have my house in view. And this threat would have to, I don't know, run CV on the feed from my camera to see when I leave or when my car leaves, then figure out if everyone in the house left so they could break in?

For this to be worth doing with hacked cameras instead of in person, you need to be monitoring a lot of camera feeds. But to get any useful info out of it, you need to spend a lot of time putting together data about a single feed. And unless you've already evaluated the target, you don't know which feeds are from valuable houses.

I just don't buy this as a real threat for any but the most valuable targets.