Yes, used it in 2017 for a fairly small dataset (order of GBs) and 5-7 nodes running in Kubernetes.
> What do you like about its killer features?
Easy admin and good at scaling horizontally.
> How interchangeable with Postgres was it?
Wire compatibility helps but there were many minor differences at the syntax level. I tried for the longest time possible to maintain compatibility with Postgres syntax and it was painful.
> How is performance?
Back then it was garbage, probably 10-100x slower than Postgres. This was pre-2.0, mind you, I heard that things improved noticeably since then. And most of the problems can be circumvented by not doing stupid things and optimizing your queries/data model. You might want to throw a cache in front of it anyway.