sqlite-wal

To prevent SQLite data loss during backups, enforce `PRAGMA journal_mode = DELETE` instead of WAL mode. WAL relies on sidecar files that can be lost during copies, causing silent corruption. Enforcing DELETE mode keeps all journaling within the main database file.

Missing writes after a SQLite backup restore typically happen because the database was in WAL mode. If the required `-wal` sidecar file is not copied alongside the `.db` file, recent transactions logged there are silently lost. Enforcing DELETE journaling prevents this.

You can detect if a SQLite database is in WAL mode by inspecting its header bytes, specifically checking the `write_v` and `read_v` values. If these header bytes indicate WAL mode, you should run a checkpoint and downgrade procedure to revert to DELETE mode.

The best way to downgrade a SQLite database from WAL mode is to run a checkpoint and downgrade recovery workflow using a script like `walfix`. This process safely merges WAL data back into the main database file and enforces `PRAGMA journal_mode = DELETE` without losing authored data.

When reviewing new sibling-tool integrations touching a shared SQLite database, verify they explicitly enforce `PRAGMA journal_mode = DELETE`. Any tool defaulting to WAL mode risks corrupting the database if sidecar files are missed during copies or incident recovery.

You should not use SQLite WAL mode if your database undergoes frequent copies, backup restores, or incident recovery operations. Because WAL separates transaction logs into sidecar files, it creates a risk of silent data loss if those files are missing during file-level operations.