The case for SQLite starts with operational honesty: if you do not need multi-writer consensus, a file-backed database removes a network hop.
Production patterns: WAL mode, careful backup/replication (e.g. continuous shipping), and clear rules for where the primary lives.
Why it matters: many SaaS features are single-tenant or region-pinned and overpay for distributed SQL.
Tools mentioned: SQLite, Litestream-style replication, and ORMs that treat SQLite as a first-class target.
Apply this by identifying services with a single writer and predictable growth, then prototyping SQLite with restore drills before cutting over.