Stale monitors and the Unknown state

Up, Down, and Degraded all describe answers: the check ran, and this is what it found. Unknown — the grey state — means something different: no answer arrived at all. UpAll marks a monitor Unknown, or stale, when no fresh check result has been recorded for three times its check interval, plus 30 seconds. All the states are listed in monitor statuses explained.

Why that formula

A 1-minute monitor goes grey after 3.5 minutes of silence; a 5-minute monitor after 15.5. One missed cycle is ordinary scheduling jitter and shouldn't trigger anything. Three missed cycles in a row means results have genuinely stopped flowing, and the extra 30 seconds keeps a borderline monitor from flickering in and out of grey at exactly the threshold.

What staleness usually means

The key insight: a failing target still produces results — failed checks, recorded with their errors, opening incidents on the way. Staleness means the measurement stopped, which points at a different family of causes:

• the probe pipeline — the machinery that runs and records checks — has stalled;
• a network problem is preventing checks from completing or being recorded;
• infrastructure has been suspended or shut down without its monitor being paused.

That's why the right reaction to grey isn't "restart the client's server" — it's "find out why checks stopped happening".

Two neighbours to keep distinct. Paused is grey's deliberate cousin: you switched Enabled off, and UpAll expects no results (pausing monitors). Down means checks are running fine and the target is failing them (how incidents open and close).

The once-per-episode digest

When monitors go stale, staff get a single digest email listing them — once per episode, not one message per monitor per cycle. When fresh results resume, the episode ends and the monitors return to their real status; if staleness strikes again later, that's a new episode and a new digest. The alert machinery itself is described in email alerts.

Why stale never counts as up

UpAll never records time it didn't measure as healthy time. If silent periods counted as up, a stalled pipeline would quietly inflate every uptime number it touched — and you'd be reporting availability you have no evidence for. Stale time is a measurement gap: visible on the status wall, honest in the record. That's inconvenient in the moment, and it is exactly what makes the numbers defensible later.