I basically do that, though I just use one xtdb; I wouldn't for production, but I'm prototyping lots of different things (throwing them at the wall and seeing what sticks). The features and use-whatever patterns make it trivial to treat them as separate concerns; that and namespacing my xtdb keys. The only real monolith is schema.clj, but you could split that up easily enough, I'm sure. If it's a matter of having different domains and the cleanest possible urls, you could do some rewriting with an nginx reverse-proxy