Sketchy thoughts about CQRS

Eventual consistency vs immediate consistency. CQRS seems to be associated with eventually consistent read models. I'm in a situation where I think I'm going to need show the user the effect of the command immediately. What's the best way to do that?

Where do the models live? CQRS is often associated with different databases, maybe completely different micro-services. In a legacy application that is a lot of effort that needs to be justified. Two different tables in the same database feel like enough to start.

There is a danger of not really understanding what needs to be in the read model. Unless you are refactoring a bunch of complicated joins, then you have probably have a pretty good idea of what the read model should be.

CQRS too early can be problematic, but can it ever be too late? Are the joins ever so complex that it's best to just rewrite everything?

I'm thinking about firing a CQRS tracer bullet. I think it could be a useful tool to master.