04 Network Latency +500ms

Time to first impact

T+0 to 45s. first impact at 0.5s.

What was injected

Network Latency +500ms — nothing was injected: this is the worked example

What happened

retries unbounded; p99 4.2s for 45s

Impact

Remediation

• Put a timeout shorter than the caller's own on the slow hop, retry at most once with backoff and jitter, and let a circuit breaker stop calls it cannot answer in time.
• Cap retries and add jitter: they were still above their baseline once the fault was gone.