Managed Service Upgrade Window
The period during which a provider may apply patches or version upgrades to a managed service, usually involving a failover or brief unavailability.
Managed does not mean unchanging. Providers apply security patches, minor version upgrades and occasional forced major upgrades, and these are scheduled into a maintenance window you choose but cannot decline indefinitely.
What this means in practice. A multi-AZ managed database applies patches by failing over, so a maintenance event is a connection interruption of seconds to a minute. Applications that do not handle reconnection cleanly — no retry, no connection validation, a pool that keeps dead connections — experience this as an outage, and it recurs monthly.
Major version upgrades are more consequential: eventual end-of-support forces them, they can carry behavioural and performance differences, and the provider's deadline is not negotiable at short notice.
The design responses. Retry and reconnect properly, and test it by deliberately failing over — this is the single most valuable thing, and it is cheap. Choose the window to align with low traffic, rather than accepting the default. Track end-of-support dates for every managed service version in use, since forced upgrades arrive with lead times that are generous only if you noticed. And test major upgrades on a restored copy before the window arrives.
The general point: managed removes the work of performing upgrades, not the work of surviving them.