Question

Support tickets report customers charged twice for the same order over the last hour. Dashboards: the payment-service consumer that processes a Kafka `order.placed` topic shows its rebalance count jumped (3 rebalances in 30 min), processing latency per message rose near the `max.poll.interval.ms` ceiling, and `charge_created` counts in your payment provider are ~1.8x the `order.placed` counts for the window. No code deploy happened; a noisy-neighbor caused GC pauses on two consumer pods earlier. How do you confirm the cause, stop the double-charges now, and reconcile the customers who were already charged twice?