[0xCMP]

I would agree, but I think it's worth checking out one of Siver's postgres repos where I thought he tested things pretty well[0]. Also things like Postgres are often run as a "server" on local dev or CI but could easily be run more ephemerally[1]. That should help with testing various states and behavior of the database.

I still would actually follow what you said thought because often times places which rely on stored procedures are, as you say, ruled by developers with the database access to manage those procedures and from what I've seen often do not have anything close to a reproducible setup. Besides the fact that often it's some proprietary database which is much more difficult to run locally than sqlite3 or PostgreSQL.

Still I keep that separate from considering if views/stored procedures are actually a good solution. I think sometimes it really is. Honestly Sivers' experiments really convinced me of that.

[0]: https://github.com/sivers/store

[1]: https://jamey.thesharps.us/2019/05/29/per-project-postgres/