I have asked Trinnov about the feasibility of this arraying beds, and their answer is clear and conclusive. In short: Arraying of bed channels is not possible with consumer Atmos. This is not a limitation of the decoder but of the way it is encoded. The spatial coding mechanism merges bed and objects into dynamic objects. There is no way to differentiate one from the other from the encoded content, the information simply is lost.Thus, the Atmos decoder can identify which objects represent bed channels and which do not, and could, theoretically, distribute bed audio across an array if it wished.
Yes, there could be a second object flying through that very same canonical position, and it would not want to be “arrayed” at that moment then revert back to point-source rendering as it moves on. Is the decoder smart enough to avoid that? Maybe – it already sees a static object at that position, so all others could either be flagged as dynamic objects or inferred to be so. That level of detail is not mentioned.