We've got some legal news to talk about.
I agree that there may be differences between the versions, but I'd counter that changes made to a relevant and responsive document don't (or rarely) render it irrelevant or non-responsive.
The burden should be on the party who altered the evidence to demonstrate why a previously relevant and responsive document has changed its character by subsequent modification sufficiently so as to lose its relevance. The upshot is that failure to track versions shouldn't enable a party to shield or withhold the "modern attachment" from production as part of the family. It was/is part of the family. Instead, producing parties who believe linked attachments have been changed are obliged to produce the changed version with the disclosure that it may not reflect the content or metadata as it existed at the time the link was created. In my view, the fact of collaboration tends not to impact the discoverability of non-privileged material. Looking at it from the perspective of the requesting party, why should a relevant. non-privileged and responsive item be withheld from production because someone altered it in the usual course of business? Where's the fairness in that?
You wrote: "The reason I think that is because the standard collection process will collect the file that exists at that link on the day we run the collection. That is not the file that was shared in most cases. We shouldn’t treat it the same as an email attachment where we would have a copy of the exact file that was shared."
I agree they are different because one embeds the information in the message and the other links to the stored message; but, I question your assertion that the link file "is not the file that was shared in most cases." Your experience is broad, no doubt, but my experience is that the linked item tends to be the same "in most cases," and that subsequent material amendments (or really any changes at all to the linked item, are the exception. Can you supply any metrics that support the unqualified assertion that "most" linked documents are (materially) different at the time of collection versus at the time they are linked? I push back because I regard you as a top expert, so I hold your assertions to a high standard of accuracy. Thanks.