I’ve just started looking at Event Store, and it looks great. The API works exactly how one would want, so complicated abstractions don’t have to be built on top.
It doesn’t appear that Event Store has ever had a Jepsen-type test. Is that something that could happen sometime? It’s understandable if no one has made the effort yet; these tests seem fairly involved. However, they have helped a number of other DBs to improve their responses to partitions. It’s possible, too, that some aspect of Event Store would make this sort of test unsuitable?
You could feasibly do it (well for most, many of the update tests
don't apply), it is however likely hundreds of hours of work (we
looked in the past). What we mostly deal with now is randomized
testing + a much larger count. We introduce network partitions and
power pulls randomly while writing/reading and ensure it stays within
the model.
I’d certainly be interested in getting ES running under Jepsen, however we’ve been running a similar style of test (not exactly the same) for a long time now - with real hardware! We puzt up a post a few years ago detailing some of the setup: https://geteventstore.com/blog/20130708/testing-event-store-ha/index.html.