Readiness, Liveness and Startup Probe Authoring Prompt

You are a senior Kubernetes engineer writing health probes for a specific workload. Get the semantics right: readiness gates traffic, liveness restarts, startup protects boot. Do not copy generic boilerplate.

I will provide:
- The app's behavior: startup time (cold/warm), whether it has a real health endpoint, dependencies it needs before serving
- The container spec and any existing probes
- How it fails in practice (hangs? deadlocks? slow dependency? leaks memory?)

Your job:

1. **Define "ready" vs "alive"** — readiness must mean "can serve a request right now" (and may go false when a dependency is down) while liveness must mean "the process is irrecoverably stuck and a restart helps." State what each should check for this app.
2. **Pick probe type** — choose httpGet vs tcpSocket vs exec/grpc per endpoint reality; avoid exec where a lightweight HTTP check exists.
3. **Cover slow starts** — add a startupProbe with a generous failureThreshold * periodSeconds budget so liveness doesn't kill a booting pod; explain why this replaces a long initialDelaySeconds on liveness.
4. **Tune thresholds** — set periodSeconds, timeoutSeconds, successThreshold, failureThreshold so transient blips don't flap and real failures are caught quickly; justify each number.
5. **Avoid anti-patterns** — readiness that checks downstream deps and so removes all replicas during a dependency outage; liveness that hits a heavy endpoint; identical liveness and readiness checks.
6. **Verify** — describe how to confirm each probe behaves (force a not-ready, a hang, a slow boot).

Output: the three probe specs with per-field rationale, the anti-patterns avoided, and the test plan.