Failing together usually only make sense for the simplest of apps, not for anything that has a lot of moving parts. If my pub-sub message queue falls over, I'd rather the web workers still stay up so visitors can see the site - they'll just be without realtime notifications. If the background workers die, those tasks should stay on the queue, but everything else still runs as normal.
So really, the math works out such that if you fail together, you'll have X amount of downtime. If you fail seperately, you'll have X*3 amount of degraded service.
This assumes that only one of X parts is essential and the others can just fail and you only get some degraded service -- which is not always (or even often) the case.
E.g if the DB fails, you're down. If the web server fails, you're down. If you're a photo service and the file storage service fails, you're down.
So really, the math works out such that if you fail together, you'll have X amount of downtime. If you fail seperately, you'll have X*3 amount of degraded service.