I still have a few times per month mails from AWS with an "instance retirement", for which then you have to plan accordingly (basically by not having a single i

I still have a few times per month mails from AWS with an "instance retirement", for which then you have to plan accordingly (basically by not having a single instance as a single point of failure). If you do the same with Hetzner, you would not notice the failure either (failed disk on a single machine? just business as usual, it will eventually be replaced). Obviously replacing AWS features like RDS multi-AZ masters is not going to be as easy and might be worth paying the whole AWS premium, but that really depends on the business size, traffic, internal experience and many other factors.