Explain why requested replicas have not cleared the queue.

All data is fictional. Four ready Kubernetes workers process a durable queue. Each ready worker sustains 50 jobs/second at the agreed safe operating point. Jobs have equal cost; arrivals and completion capacity are constant between events; there are no retries or failures. The queue is empty at 12:00:00 UTC, when arrivals rise from 150 to 300 jobs/second. Shared dependencies were tested up to 400 jobs/second, not beyond.

HPA: CPU target 60%; minReplicas 4; maxReplicas 12
Observed CPU after arrival increase: 30%; HPA desiredReplicas: 4
Namespace quota: at most 6 nonterminal Pods; only these workers use it
12:01:00  Authorized temporary scaling policy requests 8 replicas
12:01:10  Two new Pods scheduled; additional creates rejected by quota
12:02:40  Those two Pods finish 90-second warm-up and become ready
At 12:02:40: desired=8, ready=6, queue still nonempty

Assume the temporary scaling policy holds desired=8 so the CPU controller does not overwrite it. Compute accumulated backlog at 12:02:40 and whether six workers will drain it. For a conditional forecast, an approved quota increase permits the remaining two Pods to schedule at 12:03:10 and become ready at 12:04:40; all other assumptions stay true. Explain what to verify before relying on that forecast.