Docker and Kubernetes are kind of funny things because they do work poorly, and are sort of designed (if you can call it “designed”) to work that way, because failure is what they expect.

So it’s self-realizing in a way. They expect failure, are designed to fail, and do fail a lot.

I prefer designing something actually resilient, but this is what happens when you throw developers at infrastructure: you end up with nearly-incomprehensible complexity to which the only solution is more code and more infrastructure.

Which, of course, the cloud providers love and very actively encourage.