LiDar scanning works through light reflectance to measure the distance an object lies from the camera in order to build a point cloud and through connecting these points via vertices you get a polygonal mesh model as a virtual object. This method tends to perform better than photogrammetry which assembles a model of an object by detecting perceived depth from a series of photographs but bth methods often struggle creating warped or lumpy meshes or those with gaps or holes.
The iPad Pro Gen 3 has an additional sensor just below the camera flash that allows it to detect distances which is also how the built in measure app can be used as an accurate ruler to gauge an objects size. 
Reflectance of shiny materials tends to be the greatest adversary of LiDAR scanning, misplaced points or overlapping anchors can also distort these models as demonstrated below:
These models were created using the 3D Scanner App which shows the surrounding environment in red and slowly maps the object in the foreground as an array of connected points.
https://apps.apple.com/us/app/3d-scanner-app/id1419913995
I found that a sweeping movement walking around the area multiple times usually resulted in the best scans. Too many passes can result in overlapping points creating inlaid layers of a single object. It is also to be noted that the scanner on the iPad does not seem to project very far so scanning the face of a building or high ceiling was not possible. The limit this app is capable of detecting seems to be a single story in height which could hypothetically be corrected by using resources like a scaffold or ladder to bring the scanner closer to the desired area to be scanned. As expected glass and other shiny materials posed a challenge often resulting in holes or gaps in the model. Small objects less than the size of your palm were also poorly resolved. Its possible that both of these issues may be corrected by using a different application, more testing is necessary to determine the limitations of the actual scanner.