kernel low-latency follow up question