When should a service with users in every time zone schedule its maintenance window?

Open question

Guidance on maintenance windows assumes a quiet hour, and providers sidestep the choice: the Amazon RDS documentation states that the 30-minute maintenance window is selected at random from an 8-hour block of time per region, that an instance created without one gets a window on a randomly selected day of the week, and that the window "should fall at the time of lowest usage". For a service with global traffic there is no quiet hour, only quieter ones, and the hour with the least traffic is often the hour with the fewest staff awake to handle a failed change. Options a small team can choose between: the global traffic minimum; the team's own working hours with a status-page notice; windows rotating by region so no region is always hit; or removing the need for windows through read-only mode and zero-downtime migrations, at the cost of engineering work. Which of these have teams actually used, for how long, and what did the choice cost in complaints, failed changes or staff hours?

What a useful answer contains

The service's traffic distribution across regions (even roughly: share of requests per continent at the chosen hour); team size, locations and whether anyone is on call outside working hours; the rule chosen and how long it was used; how the window was announced and in which time zone; support tickets or complaints around windows compared with before the rule; how a failed change during the window was handled and how long recovery took when the owner was asleep; and whether the rule survived that event or was changed afterwards. Answers that compare two rules used by the same team are the most useful, because the traffic and the users stay the same.