I'm not sure this is exactly a new problem caused by the electronification of catalogues and the merging of what were once discrete archive catalogues into general library ones.
I can remember going back years people doing really ghastly archive reference fail - I think possibly the worst instance was somebody's thesis which was based upon, among other things, research in papers in [repository] - no indication of What Actual Collection/s in the repository in question. I would have failed the bloody thing for that alone. Though I seem to recall this in an actual published book, though I think that if one got as far as look at the actual footnotes they were a bit more informative. (Yes, I see that I moaned about that here.)
There was the letter cited to [massive and complex collection with no specific file reference] in journal article. This had not been picked up on by peer reviewers or anyone in the editorial process.
There have been various thematic online digital projects using archival materials failing to provide details of where the originals are and the necessary contextual information - I thought I'd posted somewhere about this but maybe I am only remembering various vent-y conversations on the subject. Update: Yes, it's here: I'd failed to give it an appropriate label. I've also updated a link within the post where the originating site had changed the URL.
What seems to be getting lost is the sense of the archive collection as a discrete entity and the item as embedded in a context within it - I did manage to find this in an old post on the rise of machine cataloguing on the old Wellcome blog by a former colleague:
However, an on-line database with keyword searching means that we can
no longer assume a reader starting at page one and reading (or at least
skimming) the whole catalogue: instead, a hit in a search can bring
them into the collection anywhere, without context.
Yet context matters in making sense of brief, one-line file
descriptions. If, for example, Professor X works in cancer research
funded by grants from the major charities, and then much later, in
retirement, suffers from cancer herself, a file titled “Macmillan Cancer
Care” will imply very different contents depending on where in her
career its date places it, but a reader who comes to the record without
an overview of her career, an overview of the kind set out in a
catalogue’s introduction, will not be able to make the appropriate
inferences.
Not that one didn't encounter some similar misapprehensions even in The Old Days.
I am feeling exercised about this at the moment because I was trying to find out some information about a collection of personal papers in an antipodean institution which I had actually consulted something like 30 years ago. And searching on the online catalogue of the institution threw up: certain files which had been digitised; publications by the person; the (presumably) original files; publications about person - no way I could find of getting an overview of the actual collection and what it consisted of and I am not sure it all matched my recollections.
I also feel - this was the subject of musings in another venue some weeks ago - that a lot is lost without the capacity to browse a catalogue. To find the things one wasn't actually looking for that are of relevance. Searching can only get one so far. And supposes that one is looking for the right thing (word, name, correct spelling, historically accurate term, etc).