The bottom margin

[voynichbombe]

Manuscript layout analysis is a popular and quite successful computational approach in digital humanities. There is a number of open source softwares and freely avialable services (alas the very promising Uni Hamburg AMAP Workbench seems to never have taken on wings, but the paper gives a good outline of available approaches). It is important the software is tied in with IIIF & TEIP5 capable sources so there is actually a way for classification.
I happen to use Transcribus.eu for a different transcription purpose but haven't thought of subjecting the VMs to it yet, since it's encoding is based on word level, which seems unsuitable for the VMs. But layout analysis is actually a separate task, so I will give it a try.
Generally I think that while it appears there is layout in the VMs, on a closer look it will prove to be more of a "loose analogy", like so many things VMs.
Codicology basics tell us that manuscript space was almost never wasted, but blanks filled with notes & the like, except in herbals, where even versos would be left blank.
In a parallel copying process (which is IMHO an important question for the VMs regarding speed) where quires would be separately handed to multiple scribes, text length inconsistencies would lead to either compression or expansion of writing density observable towards the endings.
All of this does not happen in the VMs. Regarded in the light of "line & page as functional units" it should make us wonder if paragraphs have a wholly different function, too. In our modern understanding it would render them as separated blocks of lists, and we know that breaking lists over pages is rather impractical.
And then there is no columnization in the VMs, and quire numbering was outdatd by 2 centuries for our ugly duckling's timeframe, so I'm afraid that in our case standards are again not really applicable for classification.