Nothing new here except the scale of the activity and the total amount of coverage over the area of interest (possibly the availability as well).

It does look like the classic problem of reflective surfaces showed up in one area of a wall of windows. The image texture shows in the right place, but with the wrong information. The lidar data tends to get really nasty on reflective things and it takes lots of samples to smooth it out.

I wonder how long it took them to acquire this and if it was ground-based or air-based. Back in 2000 I interviewed with Cyra and they gave me a demo of their lidar mapping system. Just a few thousand points per second but with a pretty good range accuracy (it was a time of flight measurement device and they were getting a few picoseconds resolution or so). It didn't work past 100m or so. To scan a whole scene took an hour or so.

Stereo cameras work fairly well in most cases but are limited in their depth resolution for a single pair. Now when you start moving the camera with good position and pose data then you can get good results.