Latency is a distribution, not a number
What repeated probes reveal about percentiles, route changes, and quiet tail latency.
Observation log
Field notes
Working notes from the edge: measurements, failure modes, and the decisions that remain useful after the dashboard is closed.
Designing a boring TLS rollout
A checklist for certificate renewal, protocol changes, and observability before cutover.
When an edge node reports healthy
Why process liveness and request quality should remain separate signals.
Small probes, useful answers
Building a compact measurement loop without adding another control plane.