There are a couple of issues with the script logic. The first is the use of the replace function in the update criteria for updateMatView - this is not generally safe. There are also some checks for internal materialization that look for a null table in the system table, but it's actually the schema that will be null.