4) is same as three but you take a subset of the data with correct relations, anonymise it carefully and use that, it has a lot of complexity at scale but is ab

4) is same as three but you take a subset of the data with correct relations, anonymise it carefully and use that, it has a lot of complexity at scale but is about as faithful as I’ve seen, it’s also one of those tools that forever requires maintenance as the core db mutates. For local devs, seeding the database with plausible correct data works pretty well.