It's interesting that they bring up the CAP theorem, because they've chosen to go for the two prongs of it that generally not preferred. Data consistency in my opinion is of utmost importance.

Though I may add it still looks pretty interesting

PM on KV here.

We want to be really clear about the guarantees here, because we want to make sure nobody is surprised by them. You’re right that this combo is preferred less for primary data stores, but we feel that means we should talk about it more, not less, so that people can properly evaluate if this is the right store for them.

There are still a lot of use cases where this is okay, which is why we built the thing in the first place.

Fair enough, thank you for the transparency then.