Documentation
Probes & regions
Who actually runs the measurements, and why more than one of them matters.
Probes
A probe is an independent process that enrols with the control plane, receives an assignment set, and runs checks on its own schedule. Each probe holds its own Ed25519 identity and signs every batch it uploads — so an observation is attributable to a specific observer, not just to “the platform”.
Probes recurse through a co-located resolver of their own rather than depending on a public resolver. Delegation and DNSSEC checks bypass it entirely and walk the root, TLD and authoritative servers directly — a validating cache would hide exactly the broken state those checks exist to observe.
Sentinel regions
Each zone is watched steadily by a number of distinct regions set by your plan — one on Free, two on Pro, three on Business and above. Regions are chosen deterministically, so a zone keeps the same observers as the fleet changes.
Burst verification
On Business and above, a detected incident triggers a burst: additional region-diverse probes are asked to re-check the zone immediately, then again on a tightening cadence out to five minutes. An important event therefore lands with independent multi-region evidence around it rather than a single observation.
Disagreement
Witnesses do not always agree — GeoDNS, anycast and CDN rotation guarantee it. A value that any recent successful observation still sees is not treated as removed, so ordinary sampling differences do not become events. Genuine divergence is what the burst set exists to resolve.