Database failover can keep uptime but return stale data

A PostgreSQL MCP server may stay available during failover yet return incorrect answers by mixing pre- and post-failover data. The replica lag causes stale reads, and every query succeeds, masking the inconsistency. Developers must enforce consistency contracts like read-your-writes and disclose lag budgets to ensure trustworthy results.