The useful question is not whether a product has one source of truth. It is whether every kind of information has an owner and a lifecycle.
Editorial content belongs where writers can shape it, reviewers can see the diff, and the publishing workflow is supported. In this starter, that means TinaCMS and Git-backed files.
Application data has a different rhythm. It is created and queried by the product, so it lives behind a small Turso boundary with explicit configuration and repeatable setup.
Keeping those paths separate makes both of them easier to reason about.