Job Title Prompting: Role-Based Reasoning for Claude

job_title: Senior SRE | years_experience: 12 | company_tier: FAANG-scale | top_responsibilities: incident response, SLO design, capacity planning | task_or_question: Our p99 latency jumped 3x at 14:02 UTC. Dashboards look normal. What do I do?

Decision: Open a SEV-2, freeze deploys, and pull the 13:55–14:05 UTC trace diff across the gateway, cache, and primary DB. Justification: normal dashboards + a sudden p99 cliff almost always means a tail-latency source not in the default SLI set — usually a cache miss storm, a connection-pool exhaustion, or a noisy-neighbor on a shared node. My default check order: 1) gateway RED metrics by route, 2) cache hit rate + eviction rate, 3) DB connection wait times, 4) p99 per-pod to find a noisy node. Missing info I need: which region, whether traffic volume also changed, and the last deploy timestamp. Next hour: freeze deploys, page the on-call DBE, and pull the gateway trace diff — if p99 is concentrated on one pod, drain it.