"My managed database has point-in-time recovery" won't save you from yourself.

RDS, Cloud SQL, managed Postgres — people treat the provider's PITR as a get-out-of-jail card. It recovers from infrastructure failure beautifully. It does NOT protect you when your own account is compromised, your billing lapses, or you fat-finger a delete of the instance — automated backups often die WITH the instance. Provider recovery lives inside the provider's blast radius. You still need an independent logical dump pulled OUT of that account, into storage they don't control. In-platform recovery is not the same as a copy you own. Fight me in the comments.