Very interesting to read this. I've had the exact same discussion a few months back with a collaborator and brought forward the exact same arguments that you posted here. The answer I got was "Yes, you are right. But things become different when you think beyond the XQuery Update Facility." Some implementations already provide 'scripting extensions', which make node identity after updates observable. I guess the XQuery Update Facility just wants to be ready for whenever such extensions find their way into W3C recommendations.