Running microservices on Hetzner is a risky move.

Their direct internet connections rarely go down, but links between servers in their internal network suffer from intermittent failures. if you make your service reliable enough to be able to run on a single node, you could have built a monolith in the first place.

If you build microservices you should always assume that links go down! So what is the deal here? Think about it as a feature to make your application more error prone. :-)

I think this is short-sighted - there are many applications where a single node is all one needs, and this is a huge part of Hetzners user base, presumably.

A bit of anecdote from me, as a decades-long Hetzner user: I have personally felt no real impact whatsoever with their internal network suffering from intermittent failures. The downtime incurred by Hetzner admin I've experienced is measured in minutes, in my case over a 10 year period as a customer...