L. D. Reynolds, Texts and Transmissions: A survey of the Latin classics, Oxford 1983, is the first port of call for any enquiry into the transmission of any of the Latin classics. On p.200-3 is the article by R. J. Tarrant on Juvenal.
Juvenal went through a period of obscurity after his own times. Not cited by Donatus, or Jerome, he is referenced more than 70 times in the commentaries on Virgil by Servius. Some of the manuscripts include subscriptions which suggest Servius may have been connected to their rediscovery: ms. K, for instance, contains Lego ego Niceus apud M. Serbium Romae et emendavi — I, Nicaeus, read this at the house of M. Servius in Rome and corrected it, and ms. L a version of the same.
More than 500 manuscripts later than the 9th century exist. Unfortunately, by the 4th century, a considerable number of spurious lines had already found their way into many copies of the text. Difficult language was sometimes replaced by simpler expressions. The vast majority of the medieval manuscripts derive from such corrupted copies.
As a rule we tend to find that medieval manuscripts go back to a single Dark Ages exemplar, or perhaps a few. In the case of Juvenal, however, we can clearly see that two ancient families of manuscripts both gave rise to medieval children. For in addition to the majority, we have a few manuscripts which preserve a more correct and less interpolated text, although the text itself is often rather more corrupt than in the interpolated copies.
The better mss. are:
P: Montpellier 125, first quarter of the 9th century, from Lorsch. Once owned by Pierre Pithou, who used it for his edition of 1585. The Pithoeanus is the best and most important manuscript of Juvenal. It also contains Persius.
Arou.: The fragmenta Arouiensia. These are five leaves from a destroyed manuscript of the 10th century, written in Germany, and broken up to use in bindings. They are now in the Kantonsbibliothek in Aarau. It also contains scholia, which are important for several reasons. Firstly each scholion is introduced by a quotation of a few words from the text. These headwords or lemmata are themselves valuable for the authentic text. Secondly the scholion itself sometimes reflects a different version of those same words, showing that the two were put together at different times.
Sang.: St. Gall ms. 870, second quarter of the 9th century. This is a florilegium — an anthology — which contains 280 lines of Juvenal. Pp.40-326 contain the ancient scholia.
R: Paris latin. 8072, from the end of the 10th century, probably French, containing long sections of the text.
V: Vienna 107, end of the 9th century, containing book 1, line 1 – book 2, l.59 and book 3.107-5.96.
P, Arou. and Sang. are very closely related. The first two are almost identical, with the text even laid out in the same manner on the page. R and V are less reliable, and V has been much influenced by the other family.
The remaining manuscripts — hundreds of them — are hard to classify. No stemma can be constructed because cross-contamination is so general, and even geographical groupings are pretty blurred. This will not surprise any manuscript enthusiast. For heavy lumps of wood and parchment, manuscripts travel about just as much as rock groups on tour, or so it seems sometimes.
Finally there are some fragments of ancient books containing Juvenal. Two pages of a 6th century volume exist in ms. Vatican lat. 5750, with scholia, and also a portion of Persius. More pages from a different 6th century book exist in Milan in ms. Ambrosianus Cimelio 3. Finally a parchment leaf from Antinoe, ca. 500 AD, contains 49 lines of book 7. None of these fragments agrees consistently with either of the medieval groups, unfortunately.
By the last decade of the 4th century, Juvenal had been equipped with a substantial commentary, which is the source for our scholia vetera (there are also Carolingian scholia), found in the three mss. P, Arou. and Sang. Mommsen discussed the date of the commentary in his Gesammelte Schriften 7 (1909), p.509-11: Zeitalter des Scholiasten Juvenals. The scholia must post-date 352-3, since there is a reference in the scholion on Juvenal book 10, l.24 to a praefectus urbis named Cerealis. But much of the material must be older, or so the footnote says. It can hardly date later than the abolition of paganism — the scholiast shows little knowledge of Christianity, and resorts to quoting Tacitus. It is difficult to believe that the compulsory state religion could be unknown in the 5th century, and indeed the writer says that the gods are still worshipped. The festival of the Matronalia is a state festival, as it still is shown in the Chronography of 354, but not in that of 449. Likewise the term used for the silver coinage is not the silliqua of the 5th century, but the older terms argenteolus or nummus.
Mommsen concludes that the commentary was composed ca. 400 AD, and that later, as is usual with ancient commentaries, it was pillaged for the materials to create the scholia in the margins of the new-fangled codex-style books.