Knex.js connection pool cleanup timing during asynchronous test execution

Integration testing suites often utilize a separate test configuration in knexfile.js to avoid production credential exposure. To ensure environment isolation and prevent resource exhaustion, knex.destroy() is typically called in a global teardown hook to close the connection pool.

In asynchronous test runners, there is uncertainty regarding the precise timing of connection termination relative to the event loop. If the pool is not fully drained before the process exits, it can lead to hanging processes or inconsistent results in parallel execution environments.

• Does knex.destroy() guarantee all active connections in the pool are terminated before the returned promise resolves?
• What is the recommended pattern to verify that no orphaned connections remain when running tests in parallel?