FounderCLI public evidence record: production database restore drill

Snapshot revision: 98ac613c18c8d10b4fa2eee57c2c64c3ecd4b9f8
Source record: docs/DEPLOY.md at the snapshot revision
Event date: 2026-08-17
Publisher: FounderCLI

What was exercised

A 33 KB PostgreSQL dump taken from FounderCLI production at 2026-08-17T01:57:35Z was restored into an empty throwaway PostgreSQL 18.4 container. The restore used psql with ON_ERROR_STOP=1 and the same database role named by the dump.

Observed results

1. The restore processed 1,084 statements and exited without an error or warning.
2. A fresh dump of the restored database matched the source dump byte for byte after removing PostgreSQL's randomized restrict and unrestrict token lines. Both normalized files had the same SHA-256, recorded in the source document as c29d9419…82868b.
3. The Lexical JSON article body in posts.content was queried from the restored database and returned readable prose.

What this does not establish

The drill did not restore over a live database, did not exercise a clean or dropped-schema restore, did not include the uploaded-media Docker volume, and did not create automated scheduling, retention, or off-machine backup. It proves that this one dump could recreate the inspected database state in an empty database; it does not prove a complete disaster-recovery system.