---
title: How alerts work
description: When we alert, when we go quiet, and how recovery is detected.
category: alerts
updated: 2026-09-09
tags: [alerts, notifications, recovery, flapping, throttled]
---

UptimeTwin alerts you when something breaks and again when it's fixed — without
burying you in noise. The goal is one clear message when a problem starts and one when
it clears.

![The Incidents page listing three incidents: an ongoing one for a firewall reading Host unreachable with Acknowledge and Close buttons, and two resolved ones showing how long each lasted and when it was acknowledged](/img/docs/app-incidents.png)
*Every alert also becomes an incident, so you keep the history: what broke, how long
it lasted, and whether anyone acknowledged it.*

## When an alert fires

A monitor is marked down only after it fails several checks in a row — how many is up
to you (see [alert thresholds](/docs/configuration/alert-thresholds)). This filters
out the momentary blips that don't mean anything, so you only hear about problems that
actually stick.

## Recovery

When the service passes checks again, the monitor flips back to up and you get a
**recovery** notification, so you always know a problem is resolved without having to
check yourself.

## No spam, but never silent

While a problem is ongoing you won't get repeated alerts, and we don't re-notify you
about an outage that is still open: one alert when a monitor goes down, one when it
recovers. On **Business** and **MSP** plans you can add a
[second contact](/docs/alerts/notification-channels), paged once if nobody has
acknowledged the incident within the delay you set.

Certificate and domain expiry are the exception, because they're a countdown rather
than an outage: while an SSL or domain monitor is inside its warning window you get
a reminder each day until it's renewed. If it does expire, that becomes an ordinary
down alert.

## Degraded, flapping and throttled

- **Degraded** — the service is up but responding much slower than its normal
  baseline. On **Pro and above** we flag it so you can look before it turns into an
  outage.
- **Flapping** — also **Pro and above**: a monitor whose *response time* keeps
  swinging between slow and normal within the hour. This is about speed, not up/down
  flips. Rather than mail you on every swing, we mute the degraded and recovered
  alerts and record a single flapping incident instead, until it settles.
- **Throttled** — the service answered, but told us we were making too many
  requests. That is not an outage: the site is up and declining *us*, so it is
  recorded for your history and **never pages you**. It mostly shows up on very
  large sites that rate-limit automated traffic — see
  [check intervals](/docs/configuration/check-intervals) for how we handle those.

## Quiet periods

To silence alerts during planned work without pausing the monitor, use a
[maintenance window](/docs/alerts/maintenance-windows).

## Get started
- [Set your alert thresholds](/docs/configuration/alert-thresholds)
- [Choose where alerts go](/docs/alerts/notification-channels)

## Related
- [Maintenance windows](/docs/alerts/maintenance-windows)
- [Making sure alert emails reach you](/docs/alerts/email-deliverability)
- [The AI assistant](/docs/advanced/ai-assistant)