When working with event sourcing, one of the biggest problem's I've encountered is protection of private data.
If a user wants their data removed, the event log must be edited - which is antithetical to the event sourcing approach.
If user data should be opaque to the system managing it, complicated encryption can be done on events, but key management becomes weird quickly.
However, if I understand this correctly, an IBM trial might give us new technology for this:
Fully homomorphic encryption could give us a way to fully encrypt personal data in events and compute on it without ever decrypting it.
We'll need to wait and see how applicable this is to domain logic outside of machine learning, but it seems like a good direction to investigate.
Personal server of Hannes Leutloff aka yeldiR