> ## Documentation Index
> Fetch the complete documentation index at: https://docs.still200.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Failure semantics

> Learn when an HTTP check opens or resolves an incident.

## Consecutive failures

Still200 opens an incident after three consecutive failed polls. A later healthy poll resets the failure count and resolves the open incident.

Failed polls include:

* connection and DNS errors
* responses that take longer than 5 seconds
* non-200 HTTP responses
* malformed JSON or an invalid response shape
* any dependency result derived as `unhealthy`

## Degraded services

A dependency with latency above 1,000 ms and no error is `degraded`. Still200 records the result, but degraded checks do not count toward the incident threshold.

## Partial failures

The worst dependency result determines the overall poll result. One unhealthy dependency makes the poll unhealthy even when every other check succeeds.

<Warning>
  Return HTTP `200` with an `error` inside the affected dependency check. Returning a non-200 response prevents Still200 from reading the structured dependency details.
</Warning>