If we decide we can't, then so be it -- things will break either way -- but it's still strange to me that we'd be okay with literally zero forward compatibility and still call that a "minor version".
Semantic versioning guidelines are not something we are following, especially here.
Our protocol version is really just two-part; just like our server major version used to be. We just happen to have named both parts here, unlike with the historical server major version.
We never have implemented a protocol change during a minor server version update, it doesn't have (though maybe it needs?) a patch version part.