Assume your organization stores order book–related data in multiple databases (e.g., one for raw exchange messages, one for normalized orders, one for trades/executions, one for analytics). You suspect data quality issues and want to validate that **all stored data follows order-book logic**. Design a validation approach that answers: 1. What **invariants** must always hold for orders, executions, and the resulting order book? 2. How do you validate **within one database** (table-level constraints, batch checks)? 3. How do you validate **across databases** (reconciliation) when they may be eventually consistent? 4. How do you handle edge cases: partial fills, cancels, replaces, out-of-order events, duplicates, and reconnect replays? Be explicit about what you would compute and compare, and how you’d operationalize it (alerts, dashboards, backfills).