Major versions don't come out suddenly. We have time to test and make sure they work well with our storage.
Customers don't always want to upgrade to new version when old versions just work. However this can lead to version creep. Ideally we want to always run the latest version of Postgres. We will hold this line as long as possible.
Sorry, I meant less "how long until they catch up to trunk" and more "what is the downtime / operational burden of performing major upgrades like".
With separate storage and compute it seems tantalizingly possible to have low-effort zero downtime upgrades for the user, which would be a huge selling point.
Customers don't always want to upgrade to new version when old versions just work. However this can lead to version creep. Ideally we want to always run the latest version of Postgres. We will hold this line as long as possible.