This is frustrating, how careless teams are to just copy the production data. I have gone deep into this and most reason why developers go for this is that the fake data immediately available to them are not always in the same shape for them to reproduce the bugs, fro eg. relationships, skew and edge cases. Anonymisation are just a wrapper fix, because skilled hackers could deanomalise these datas accurately.

Because for medical records especially, quasi-identifiers such as dates and postcodes can be utilised to re-identify patients.

As a solo developer who is working on generated synthetic data with features to declare distribution and patterns along with checked foreign keys, I feel responsible to build my software better and distribute it to all teams that need realistic data for testing their production systems.

Atleast scoping the test bucket would have been a cheaper fix.