My issue with this is that traffic cameras are not necessarily independently verifiable by a third party observer unless they have physical presence or a camera of their own. At least with the weather you have multiple independent sources verifying the data. Satellite imagery from multiple states or organisations and local weather station data by national organisations provide a pretty impartial result.
I agree with your main concern. But traffic betting cannot be reduced to just any fake livestream or imaginary place problem. This does not address the real oracle problem.
This type of betting is already a reality today. There is betting game of counting cars from live CCTV. Where the operator uses AI to count vehicles and use that as the result. But there is really no way to verify what we see on the screen. So we need a verification system where even if there is manipulation in one place, another data source can catch it. Like- multiple independent feeds, immutable timestamped recording, sensor provenance, anomaly detection or predetermined fallback sources etc.
And weather-station manipulation is really concerning. After an unusual temperature spike in sensor at Paris Charles de Gaulle, Météo-France filed a police complaint. Same time, some traders at Polymarket made big profit. Since investigation is still ongoing into the alleged tampering, it would not be right to call it proven cheating.