The above statement, by Henry Bradshaw, could be the official Creed of the provenance researcher.
When discussing cataloguing, I sometimes tell people that I think that the single most important thing a cataloguer can tell their reader about a manuscript is its collation. (This is partly, but not exclusively, because the physical structure of leaves and bifolia is one of the few features of a manuscript that usually cannot be conveyed by photography). People who study manuscripts -- and early printed books -- seem to fall into two camps: on one hand there are those who see collation as a mechanical task, like measuring the leaves and counting the number of lines, and on the other hand are those who really understand how much a book's physical structure can tell them about its manufacture, and know how to use this powerful tool to inform their work. I won't go into details here; I mention it only as background to explain why Henry Bradshaw is one of my bibliographical heroes -- on a par with M.R. James and Neil Ker.
Last summer Maria Saffiotti Dale, Curator at the Chazen Museum of Art, sent me just the sort of provenance puzzle I enjoy. It concerned a 15th-century Italian cutting, shown above, described on the Chazen website as an "Initial 'D' from an Antiphonary from Como Cathedral with a Temple". 
On the back is an inscription which she could not entirely read: "I can make out all the inscriptions EXCEPT the initials(?) immediately preceding the date 1856 and what follows it. Any ideas?"
The Schoenberg Institute at the Kislak Center at the University of Pennsylvania is the only institution, as far as I'm aware, that regularly introduces its medieval manuscripts by way of YouTube videos; there are now more than 70 (including non-western MSS). I only recently became aware that one of these videos concerns not a codex, but two collages of cuttings, perhaps assembled in the 19th century, one of which is shown above.
This caught my attention because I had also recently been looking at other fragments from the same manuscript.