How do I check reliability?

Short explanation:

Reliability means the operation behaves predictably when dependencies fail, time out or return uncertain outcomes. It includes recovery after a crash, bounded resource use and preservation of business state. A happy-path test cannot prove that retries, rollback and cleanup are correct.

Key points:

• Follow timeout, retry, crash and partial-failure paths.

• Check transaction boundaries, cleanup and duplicate processing.

• An external action may succeed even if its response is lost.

• Trace what is durable before and after each external effect, and check whether a retry can repeat that effect.

Interview point:

• Identify a specific failure window and the state a later retry or operator will observe.