We did this at Verizon Smart Communities, migrating our Neo4J customer/vendor database to AWS Aurora MySQL. Neo4J is very good at representing parent/child/leaf relationships, but we did the same by creating a table with node data, and a reference to a parent ID, including indexes on both the row ID and the parent ID. This did require code changes, and some complex queries, but once developers understood that it's easy to represent graph data in a relational database, the changes were made, query performance improved dramatically and the system has been outstandingly reliable at scale.
