Trying to see what I can do with the relational model, only supporting multi-model. Tagging each row of tables with a “model” identifier, much as temporal databases attach an interval or whatever to a row. This would be give us a First Order Modal logic. I could have some fund pondering what various modal logics can do here (I’ve studied Modal Logic, but not recently). It occurred to me to ask about existing work in the area. At least what the concept is called, so I can go look for prior art. Think of having a Datalog database and being able to set a “current model” or being able to do separate parts of queries in different models. I’m also interested in the idea of merging multiple models to build new ones. What should I look for? Thanks.

You could look at Deductive Databases, but that's advice from the 1980s for you. Marek Sergot was doing that at Imperial when I was there. Obvs, there's the Semantic Web stuff and I'm sure there are DBs for that. Can you elaborate on this "model" field? Maybe an example.

The model is an opaque value. Queries could be restricted to rows where the model has a particular value. It could be used for multi-tenanting, for example. One could use multiple models together in various ways. Branch-merge change tracking. Multiplayer services of various kinds could be seen as merging the separate models of the different users’ changes. Models could override and inherit from each other, too. Given the right primitives, all sorts of things could be implemented in userland, securely.

Hmm, so it's not like a (hash)tag because it's never a list?