(though note that on current websites, sometimes the police have fun adding random information)

If there's an accident, one way packets could reach drivers that are approaching the congestion is to hop from one car to the next, in the same lane that the accident occured. However, that requires many hops, and requires contiguous coverage (which may not be available in the early rollout of such devices). A potentially faster way is for the packet to be carried by the opposite lane (as long as they're not heavily congested as well). In this way, only one car is required to carry the information to all capable receivers, and the information reaches approaching cars at the rate of speed of the other lane (which, while noticably slower than the speed of light, is still MUCH faster than finding out about an accident without such a network).

It's interesting that such a system would mean that police officers would no longer be able to sit by the side of the road waiting for oncoming speeders, because the people in the other lane would have already alerted them of the police officer's position. (drivers already do this sometimes with their headlights. However, even when they're willing to let oncoming drivers know, they often don't want the police office to know that they're directly subverting him, so they tend not to do it very much. If the ad-hoc vehicular network allowed anonymous messages to be sent, then drivers would be free to alert others of the officer's position. (they could also do so unambiguously, without worrying that other drivers might interpret it as "deer in the road" or a wide variety of other possible signals))