Well, we haven’t had an answer, so I’ll give it to you:
All of the numeral-phrases except the bottom right read CXLVII (= 147). I’ve highlighted every other character in red to emphasize the distinct characters, making this solution comprehensible if not obvious. The fact that you, my readership (several of whom have training in paleography) couldn’t find the solution even after two clues demonstrates the ongoing importance of paleography as a scholarly profession, and also demonstrates the highly aberrant character of the hand.
This plate is taken from a 1926 article ‘The use of Arabic and Roman numerals in English archives’ by Sir Hilary Jenkinson, the doyen of British archivists (and who, shockingly, lacks a Wikipedia page!), who was also one of the major figures in early-twentieth-century British paleography (Jenkinson 1915, 1926). Jenkinson writes of this plate:
First there is the unimportant but highly curious development in one or two Courts of a special Roman numeration for the membranes of their Rolls, which in its later stages is practically unreadable. Our illustration (fig. 10) shows the number 147 as it was written in the years 1421, 1436, and 1466 and finally the number 47 as it appears in 1583: it does not show the worst that might be selected and is only half the size of the original (Jenkinson 1926: 274).
Although I have been familiar with this plate for over a decade, for the life of me I can’t find a way to read 47 (XLVII) out of the bottom-right phrase. I’m sure that Jenkinson had some grounds (probably contextual) for believing this to be the numeral, but I just can’t get 47 out of it. Weirdly enough, I can find 147 in it, but that may just be a product of its association with the other three phrases. Note that the other three are chronologically close while the fourth is nearly a century later. Anyway, I’m really stumped on this one.
But I actually wish to disagree with Jenkinson on one word: ‘unimportant’. The cursive transformation of cumulative numeral phrases is important in the history of numeration because it is the most common means by which cumulative systems, which rely on the repetition of like symbols whose values are added (e.g. XXX = 30), turn into ciphered systems, which do not do so. The figure below (borrowed from my book) shows how the Egyptian hieroglyphic numerals for 6, 9, and 300 became paleographically reduced over time into hieratic forms that do not show any evidence of their original cumulative structure. The ligaturing of individual signs eventually leads to the reconceptualization of the entire set of signs as a single unit. In fact, this process occurred (slightly differently) with the Brahmi numerals 1 through 3 of ancient India, which eventually became the ciphered figures of the Indian, Arabic, and Western numerals used by virtually everyone today (see this chart, for instance).
The Roman numerals ceased to be used in manuscript-writing throughout early modern Europe, so there was no opportunity for the form of this particular court hand to lead from extreme cursivization to something structurally distinct in the Roman numerals. But it could have happened, just as it happened before several times. And that (along with so many other things) is why paleography matters.
Chrisomalis, Stephen. 2010. Numerical Notation: A Comparative History. New York: Cambridge University Press.
Jenkinson, Hilary. 1915. Palaeography and the practical study of court hand. Cambridge: Cambridge University Press.
Jenkinson, Hilary. 1926. The use of Arabic and Roman numerals in English archives. The Antiquaries Journal 6:263-275.
Glossographia 
