In event sourcing you need to somehow pass around events, which are
arbitrary data that describe necessary change of the state. Kafka is
a message router/forwarder, with message being an arbitrary data. No go
figure where does that fit.
I think people are doing ES in different ways, and I wish I could hear about the architectures around: what kinds of internal services people have producing and consuming the events, where they process the side effects of the events, where to store the events and also how they are updating the databases used by the applications. I mentioned Kafka because it plays a major role for several implementations, but Kafka is not the whole story neither it is obvious how to deal with it.