A closing thought is that I suspect the manuscript was copied from
something like an "eight-up" rough draft. This would be something like a
bifolio-sized page containing miniature sketches of about eight pages of
illustrations, with the master text being supplied separately. This would
make sense of some of the apparent copying errors - for instance, the
plants with roots that look like nested cones might well have been
intended to have ridged tapering tuberous roots like carrots. This method
of planning is quite a bit faster than producing a full-sized draft of
each page. It also fits with the use of unique "labels" as the first word
of many pages - that makes sure that the right text goes with the right
illustration. (Though there are doubtless many other explanations which
make equally good sense!)