[thephyber]

> There are more established patterns for data collection than data destruction. How do you know when it’s safe to delete some piece of data?

I agree, but this is exactly analogous to the SDLC. Most coders only learn to hack together barely-working code. Those who spend the effort to learn the craft figure out how to {version control, unit test, static analysis, benchmark, integration test, upgrade library dependencies} and automate these processes.

Similarly, there needs to be a data lifecycle with defined retention lifetimes for different data, defined processes for actually disposing of data, and special handling cases for backup blobs (which may be retained longer than the retention lifetime of a subset of the data in the backups). This is effectively intended by the GDPR (not sure if it states explicitly) and similar laws.