A real-time endpoint has rising latency and errors during peaks. Which two investigations best distinguish capacity pressure from application failures? Select 2 answers.

1. Increase maximum instances without checking whether the failure is authorization-related.
2. Check only the training dataset row count.
3. Inspect invocation rate, model latency, errors, and relevant instance utilization.
4. Inspect container logs and downstream dependency behavior for failed requests.
5. Disable all error metrics to stabilize the dashboard.