Many of these planes simply capture high resolution panorama timelapses of a large portion of the city.
The police can then wait for a crime to happen (e.g. a burglary), and then if they have coverage of the time, they can just play their timelapse in reverse to figure out where the vehicles involved came from or went.
Then, they use "parallel construction" to erase the fact that they used persistent mass surveillance to catch the criminals. They already know whodunit, so it's just a matter of coming up with a "lucky break" or an "anonymous tip" or whatever to explain how they caught someone, without ever exposing the surveillance to the scrutiny of the court.