In #1 Array JSON was turned into JArray. I think this is a useful thing to do and a good assumption that a JArray is not necessarily an Array. However all JArray operation are implemented in terms of toArray. So we've changed the assumption from "It's always an Array" to "In order to observe anything, it must be converted to an Array". This implies that there is a 0-cost conversion to Array, since that would be the only way one could write performant codecs. Should JArray operations be FFI like JObject?