If the job is successful, it should finish without coming closes to the timeout. Extending the timeout will make it take longer to fail, but since the job is flaky, and the related code is rarely touched, it seems worth it.