From Steve Huston's comments:

Part 2, section 7.5, delivery-id: How would one know if a message with
delivery-id(2) is a mis-id'd continuation of delivery-id(1) or a new
message? By using the delivery-tag as well (if it's missing, it must be a
continuation)?

Part 2, section 7.5, settled: What is the desired/required behavior if the
values in different transfers of a multi-transfer message are different?

RS: We should make it clear that settlement in general is idempotent, i.e. things can go from unsettled to settled, but not from settled to unsettled. Probably in 2.6.12 where we first talk about settlement.

RG: Should also add text to other fields (e.g. delivery-tag, message-format) defining that they cannot change between transfer frames on the same delivery (though they may be omitted).