...
I can't actually use that feature to gather the information I'm interested in since all I have are tables and data with no formal information on key relations and allowed load order. There is so much data in so many tables that testing even a single load iteration takes many hours and there are literally many thousands of load order combinations possible. Logs of hundreds of thousands (or millions) of XML loads would take a very long time to go through, and would then only give one possible load order.
Automating a diagram of key relations and visualizing it is the first step to methodically computing a correct load order, but I can't do that until I figure out how to use the system tables to describe (1) columns which are not keys, (2) columns which are primary keys not referring to another column, and (3) columns which are foreign keys and the table/column they are pointed at. My SQL knowledge is somewhat limited and I am struggling with the system tables.