I have heard arguments on both side of this one at many IAEI meetings across the country, and while in all cases the CMP's stated the intent is to not depend upon the box connectors at cabinets and box's, the other sides will always bring up the fact that the couplings in the conduit would also have to be bonded across, as they present the same problem.
the two problems arise from the terms "physically continuous" and "electrically continuous" and with both being the requirement of bonding across for the lack of either one.
The beginning of 250.64(E) has the statement
Ferrous metal enclosures for grounding electrode conductors shall be electrically continuous from the point of attachment to cabinets or equipment to the grounding electrode
but then does a back flip and states:
Ferrous metal enclosures that are not physically continuous from cabinets or equipment to the grounding electrode shall be made electrically continuous by bonding each end of the raceway or enclosure to the grounding electrode conductor
so how do we defined "physically continuous" if it is to say any ferrous raceway that is not run as one long piece of conduit without any couplings or connections through a box, then I say that is a far stretch.
So IMO I say as per the wording Don has a case, but I also know the intent was to not depend upon the connections because of the high current that they might be subject to, same as service conductors.