Scheduling and alerts
Checks run on a fixed cadence; alerts fire on transitions. The gap between those two is deliberate.
← All docs · Updated 2026-09-04
Cadence
Run frequency is set by plan, from a 30-minute cadence on the entry tiers down to once a minute at the top. A cron-driven scheduler handles lower volumes; a queue-backed worker handles the rest. An on-demand check is always available from the dashboard and CLI for when you want to confirm a fix immediately.
State-change alerting
An alert fires when a store transitions between PASS and FAIL, not on every failing run while an outage continues. A cooldown (60 minutes by default) suppresses repeat notifications inside a single incident. UNKNOWN runs never alert.
Alerts are delivered by Slack, webhook or email. The channel carries the failure code, the affected store, and a link to the run with its full evidence.
Impact classification
Each incident is classified by who it affects: all-customers, some-customers, none, or unknown. That classification, plus a root-cause confidence, is what lets you decide at a glance whether a failure is a page-everyone emergency or a known edge case.