My concerns with cameras were dirty lenses and limited dynamic range.
The first is pretty easy to fix with more cameras, or the same technology we use to solve it for humans (wipers). LIDAR is also vulnerable to dirty lenses, both on the emitter and receiver.
The second is an imaginary problem: human eyes and modern cameras have about the same dynamic range. The human eye fakes high dynamic range the same way we do it with cameras: taking multiple exposures. And the camera can do it faster.
I think LIDAR became popular with autonomous car developers because it was off the shelf when they started and it gives a nice clear single-sensor direction + distance measurement. LIDAR's big problem is that it's serial. Waymo's LIDAR apparently has a refresh rate of about 10 Hz, which is pretty scary if you've ever played a racing game at 10 FPS. It's also mechanical, which isn't great. They also don't talk about the resolution, probably for good reason.
Automotive LIDAR's big advantages are that near IR can penetrate fog a bit better than visible, you get a depth measurement that's not dependent on the limited baseline you have in a car. But you can make a camera that's sensitive to the wavelengths used by automotive LIDAR pretty easily. You can also make time of flight cameras, again pretty easily. And those cameras are full parallel devices with no moving parts and as many FPS as you care to pay for. They also don't require shooting lasers around, although you can if you want to.