This is the maximum latency observed across the sampled test frames at this frame size, capturing the worst single delay spike seen during the test rather than a typical or average value. An admin pays particular attention to this figure for latency-sensitive services, since a single large spike (even if rare) can be enough to violate a strict voice or synchronization SLA even though the average latency looks fine. For example, an average latency of 50 microseconds alongside a max of 3000 microseconds would flag an intermittent congestion event worth investigating rather than a stable, predictable delay.