The reason for not including the responseToMessageId in the multiple messages sent in response to a single message is not clear. They are technically speaking all send in response to that message so the fact that there are several ones does not seem to be extraordinary. Moreover, I would also expect some standard mechanism allowing to quantify the number of messages to be sent in response. For example in a response specify the number of response messages expected and the index of the current response message.

Existing Wording:

Applications SHOULD be prepared to see, at most, one incoming message with a given responseToMessageId. If multiple response messages (e.g., streams) are needed, this can be accomplished by having the server send “unsolicited” messages, i.e., messages with no responseToMessageId, after a client’s initial request.