6 years ago when I used Hetzner, it was widely known as unreliable, a provider where you could get hosting that delivered good performance for a cheap price. The tradeoff always was that you needed to treat each machine almost as EC2 Spot: it could go down at any time.

You also needed to consider that when this happened, the data inside the machine was mostly lost. Finally, you also needed to plan to graduate out of it as soon as you had enough money to go either to a colocated data center or the "real cloud".

I kept Hetzner as a backup provider in more than one company, mainly to have real machines for take home tests, back when hiring was plentiful. Even so, we often faced problems with the machines going down due to hardware or networking issues, and the need to rebuild them from the ground up. Those mirrored all tales of woe everyone in the department had from years of working with Hetzner, sometimes losing production data because the rules of the game were not followed.

So it seems that 6 years later their scale has increased but the experience remains the same. On the bright side, kudos to Hetzner for teaching waves of engineers about reliability and disaster recovery during all these years.