[quectophoton]

In my experience, chances are that the database will outlive whatever application code is layered right on top of it.

So ensuring the database itself protects the data integrity and prevents the application code (current or a future refactor or rewrite) from messing it up, sounds to me like the sane thing to do. Be it with triggers, with functions, or whatever.

Though I can understand that people usually don't like how PL/pgSQL looks like (I don't). But if you ignore the ugly language syntax, testing it is no more difficult than testing, say, an AWS Lambda function that is triggered by SQS and writes stuff to DynamoDB.