Why is provenance important in research?

Because the question always arrives later

A reviewer assessing a grant application, a partner running diligence, a security team before data moves, an acquirer years afterwards: each asks a version of the same question. What was produced, by whom, and has it changed. None of them were in the room at the time, and the people who were have usually moved on.

Because reconstruction is expensive and weak

Rebuilding the account afterwards costs weeks and produces something less convincing than the record would have been, because it rests on memory and on files whose history has been flattened by renaming and overwriting.

Because the actors changed

Automated agents now run steps of research on their own, at night and in parallel, with nobody watching. Attribution used to be a matter of asking a colleague. It is now a property a system either records or loses, and losing it is only noticed at the moment somebody needs it.