> Why can't the phones determine their proximity to each other
Because GPS in built-up areas has super low precision and position estimates jump by tens of meters. A cart of phones is going to be almost indistinguishable from a pack of cars.
"But missosoup don't phones also have accelerometers and gyros?"
Sure, but they're used to feed into a Kalman filter that produces the position estimate. Apps don't generally access high resolution sensor feeds because the positioning service is already doing the best job possible. Even with that, it's hard to distinguish multiple cars accelerating from a green light compared to a bag of mobile phones being carted around.
tl;dr: it's hard. Measures can be taken to filter out these attacks but they'll come at the expense of reducing the accuracy of regular traffic data. Unless these attacks start becoming a real problem, it's probably not worth it.