A handler times out while waiting for a downstream HTTP call with no client deadline. Retries can repeat writes. Which changes are appropriate? Select 2 answers.

1. Set suitable connection and read deadlines within the invocation time budget.
2. Ignore all downstream exceptions and always return a completed result.
3. Generate a new order ID after every timeout and repeat the purchase automatically.
4. Use bounded retry policies and idempotency for operations that can be repeated after an uncertain response.
5. Remove every timeout so the function can wait indefinitely.