Thanks for your comments. Obviously I do see a distinction between a distinct protocol element that reaches some form of state synchronisation with its peer and a modification of a protocol element where there are additional behaviours added to the element, but within an overall synchronized state of an existing protocol element.
The properties you mentioned are not orthogonal that that's not a classification.
For example, how about a modification of a protocol element where there are additional behaviours added to the element not within an overall synchronized state of an existing protocol element?