Company was moving microservices and containers anyway, so Kubernetes is one of the few sane options to run them.

I can't say that anything got worse. Once you (collectively) get past the initial learning curve and add automation in place, it's way better than most other deployment scenarios. Kubernetes worker nodes fail from time to time, AWS kills them and we barely notice. POds get rescheduled automatically and mostly work.

Frankly, the remaining headache-inducing things are mostly related to the software which is not running on Kubernetes (mostly stateful infrastructure, and mostly for non-technical reasons). Managing VMs is a pain.

There is one thing that one needs to be careful. If you are moving from a VM for each service, to Kubernetes, now suddenly your service is sharing a machine with other services, which wasn't the case before. So I'd suggest that proper resource limits be set so that the scheduler can work properly.

One thing that can get worse is network debugging. There are way more moving parts, so it is not as easy to just fire up wireshark.