I had also previously used eventstore, and it was a very powerful, highly optimized event sourcing back end. I was not savvy enough to get it set up consistently on a Windows server at the time though. Perhaps it is easier now, plus I'm (slightly) more used to linux.
Interestingly, this thread has helped me realize that ibgib's current approach has properties very similar to a mix of event sourcing, neo4j, IPFS, and "functional" code-as-data. But I digress.
Both CouchDb and EventStore were great tools in my (non-production) experience.