Absent a need for asynchronous, independent access from multiple external web clients, I don't get why the author would want to do this. Isn't marshaling/unmarshaling always major infra overhead?

I get using Ruby because you like Ruby. I like Clojure, and if I was building a new backend for ordinary web stuff, that’s probably what I’d gravitate to. But it sounds like Go was working til he got fancy with the architecture.

Just because you don't know how to use Go toolset the right way, does not mean it's the wrong tool in general; but I understand your point(s).