Since 08:44 UTC, four changes to the slow / normal-again alerts, after one account received 182 of them in a night, all from sites behind Cloudflare judged from Oceania and Africa. Replayed on the same night's data, the new rules send none.
- Nearest your server, behind a CDN. When every continent connects to the site within 60 ms, there is no nearest continent, so the monitor's visitors' regions decide instead. The alert says so: «… in Europe (your visitors’ regions; the site is behind a CDN)», and the monitor form's info text explains it.
- Five minutes, not two rounds. A monitor turns slow when the median is over the threshold in every round for 5 minutes, and normal again when it stays under 80% of the threshold for 10 minutes. Between the two lines nothing changes.
- Flapping. The third time a monitor turns slow within 6 hours, one «Speed keeps changing:
» alert goes out and its speed alerts pause for 6 hours. The monitor page still shows every change («Slow · alerts paused 6 h, it keeps changing»). Webhooks get the new event monitor.slow_flapping. - One message an hour. On email, SMS, Slack, Teams, Discord, Google Chat, ntfy, Pushover and Telegram, the first speed alert of a quiet hour goes at once; later ones within the hour come together in one message an hour after the last, «Speed: 1 slow, 1 normal again», each site with its times. Each channel has a new option, Send speed alerts one by one, at once. Webhooks and PagerDuty still get one call per event.
Down and up alerts are unaffected. Details in Alert rules and Monitor types.