I’m quite experienced with protobufs and they handle it well enough. Having no schema doesn’t address the root problem of API producers/consumers needing to manage schema changes together.