After changing a configuration in one service, I’d like to notify the services, but not concurrently on all the servers at once. So if I change the settings - that require a restart of apache and supervisord - I’d like the 100 hosts slowly restart apache/supervisord, one at a time. That restart is currently a handler in a role.
what you might want to change it to have the actual handler execution
in a subsequent play, and have the current handler set the host in a
'needs restart' group (using group_by). This way the original play can
be concurrent and the new play will using this new group with serial 1
and have the task to do the restart.