Proposed solution is to simply ignore any additional keys that don't match the schema. This is effectively what we are already doing, since we simply store the additional keys in a private Map field.
This field will be serialized when written to XML by XStream. The extra keys also end up being put into the JSON string representation which, means we could be storing additional key/value entries that are inaccessible to the back-end and would only be visible to clients. That is not an intended use case for the extraFieldFromNewerVersion Map.