With the previous discussions on URDF in mind, we have created a public dataset containing URDF files from different sources.
The files were analyzed to find commonalities and discrepancies across files, and compare URDF parsers.
The results and information about the dataset can be found in the arXiv article.
I believe some of the most interesting results are in tables IX and XVI:
A multiply defined robot is a robot model that is provided by different sources, e.g., two sources may both provide a URDF model of a UR5e robot.