99.9% SLA: what it means in practice

99.9% still allows ~8.7 hours of downtime a year. Check whether planned maintenance counts, if there’s a public status page, and what compensation actually is (a hosting credit ≠ your lost revenue). Use external monitors (UptimeRobot / Better Stack) from multiple regions and keep incident logs. For e-commerce, TTFB and DB stability beat a “99.99%” banner. Who’s had real downtime — how did the provider respond, and what did you get back?