Different applications could use different schemas, you just keep in mind the eventual consistency. i.e. you can use event sourcing or just treating everythin

Different applications could use different schemas, you just keep in mind the eventual consistency. i.e. you can use event sourcing or just treating everything as a log that gets read and merged (think of it as a CRDT) should be able to handle those kind of scenarios. That use case matches Cassandra quite well with its wide rows.