Probes run every minute from the API.
The parts you notice right away when something is wrong.
What keeps the rest working. If something breaks here, it shows above.
Derived from the probes: an incident starts at the first failing probe and ends at the first one that answers again.