Certainly there are circumstances where Lidar fails but I think you are exaggerating the difference. In many cases camera fails too, it certainly can’t see through fog for instance.
If the fog is thick enough, yes, that's true. Of course, up until you lose vision completely, you can usually compensate for fog well enough to detect objects crudely by just cranking up the contrast. That won't get you the text of road signs (e.g. speed limits), but most of the time, that behavior isn't safety-critical, and when it is, the signs also typically have a distinct shape, so reading the word STOP or ARRÊT or whatever usually isn't critical to detection or interpretation.
With LIDAR, you have to use statistical techniques to guess whether a return is real or not, and that only goes so far unless you have an insane amount of data (read "scanning the same scene for several minutes"). So the jury is still out on whether any of those techniques will turn out to be practical in real-world use.
And you can do the same sort of analysis with a large number of camera photos taken over several minutes, too, which is to say that even if those techniques do prove to be viable, they still won't necessarily give LIDAR any particular advantage over cameras.
And for daylight fog, the fact that the light source (the sun) is diffuse should give cameras a large advantage over LIDAR, because you aren't generating a point source of light that gets reflected in nonuniform ways. At night, because headlights are point sources, they could theoretically be almost as much of a problem as LIDAR, though changing the beam angle (fog lights) can significantly reduce that problem.
