However, if the blue car passed a license plate reader it stops being a random blue car and becomes license plate ABC-1234 registered to someone and your now "following" that specific car plus the original subjects car for that hour. Basicly, the readers enable you to track people not just where the reader physically is but wherever you have footage either leading up to that scan or after the fact. Together you can potentally follow millions of people for fairly low costs.
Granted, nothing says the person who the car is registered to is the actual driver, but you have someone who can probably tell you who was driving that car.
I see absolutely no way that bringing up the existence of license plate scanners created a situation where there was no slippery slope.
Also the recorded video collection grows to be a ridiculously large amount of data over time, even detecting all video of John Smith's car across the United States will become an expensive task to complete computationally. And for everyone? Well even more expensive.
Add on the concept that camera's will not be covering all of the United States at times, and if you are just using make, model, and color of cars to track people between license plate readers shell games where similar cars rearrange them selves in tunnels or garages will become extremely effective at defeating surveillance if license plate readers are not basically everywhere.
So in spite of you deciding to bring up license plate readers, and explain how they can be used to assist in tracking with airborne surveillance I still fail to see the relevance of how they ensure that 'everyone' will be monitored.
I suspect if you wanted to harass your ex-wife, or a dissident, then overhead footage of their car for a few hours would provide a multitude of crimes.
Sure, if you actually prosecuted everyone that's one thing, but selective enforcement plus mass surveillance is a very powerful and easy to abuse tool and you don't need anywhere near 100% coverage to make this effective.