At least the standard doesn't seem to support that: it suggests creating the ID by hashing the current time, file path, and file size. Besides, there are a billion other side channels that can be used to digitally watermark PDFs.

Anyway, I've since found an explanation in the PDF 1.7 spec. PDF files can include "file specifications" (file paths or URLs) to refer to other files. An ID can be added to a file specification for an external PDF, which the PDF reader can validate against the ID contained in the located file's trailer dictionary, to ensure that the external PDF is the expected one.

I guess "some workflows" must have gotten upset when trying to generate a link to a PDF file with no ID, so they finally required its presence in the spec.