The Joyce manuscripts acquired by the National Library of Ireland in 2002 form a substantial new body of material that will take some time to process. The avant-texte of Ulysses has expanded and changed. The initial impression, from both Mike Groden and Daniel Ferrer, is that the new manuscripts contain little that is dramatically unexpected.  Most of the new manuscripts fill in gaps from previously-known collections, specifically the Buffalo collection, and do not, in general, depart from expectations. There are, apparently, no immediate breathtaking revelations, such as, say, the unequivocal assertion that the man in the macintosh is actually Garryowen erect (a theory I steadfastly, if unadvisedly, cling to). This does not mean that further study of these manuscripts will not eventually alter the paradigms we presently hold. The story of these manuscripts has only just begun.
Initially, the most surprising new manuscript is the early Proteus draft, which represents that episode in an unexpected configuration. Up until this new draft came to light, the earliest-known Proteus draft was Buffalo V.A.3, which represents the text of the entire episode in a form mostly congruous with the final text, with a few notable exceptions. The new NLI draft (II.ii.1) contains the text of only about one third of the final episode divided into sixteen discrete, fragmentary units, out of order from their appearance in the final text. Comparison of the NLI Proteus with the Buffalo draft shows that, for the most part, the earlier drafts units were transferred to the subsequent draft with minimal modifications. However, the Buffalo draft contains a great deal of text not indicated on the earlier NLI draft such as the famous opening passage on vision and colour and so the later draft must have been made in conjunction with another, still-missing, draft. Furthermore, there is evidence that for at least some of the units on the NLI draft, an intervening draft was also made. 
Daniel Ferrer has suggested the possibility that at this early stage Joyce might have been considering a fragmentary form for this episode, although such a claim is difficult to maintain unequivocally without further corroborating evidence.  The fragmentary disposition of this draft could easily be a temporary state. Piecemeal composition is not an unknown practice with early Ulysses drafts. For example, the earliest Cyclops draft, Buffalo V.A.8 and its new-found companion NLI II.ii.4, represent the episode in a series of vignettes bereft of links; and the earliest Circe draft, Buffalo V.A.19, represents the second half of that episode in a rudimentary, out-of-order, state. The difference with the NLI Proteus, though, is that most of the units it contains are exceptionally brief. Ferrer observes that this draft reads like Giacomo Joyce, and this comparison is instructive, even if Joyce never intended this fragmentary form to be the episodes ultimate configuration.
I would propose that Joyce, at this early stage of the development of Proteus, conceptualised the episode as a series of discrete units that would be be rearranged and linked together through additional material into one more-or-less integral narrative thread. In other words, the fragmentary arrangement was fungible. Proteus is more Protean than previously assumed. Each of the drafts units focuses on a specific scene or thought and could be construed as being analogous to Joyces concept of the epiphany. Indeed, the final unit on the draft is a revision of the Paris epiphany (epiphany 33). In a sense, then, this draft attempts a revision or remaking and remodelling of the epiphany, which had been the central component of Joyces æsthetic theory in 1904, that is, at the time of Stephens Sandymount stroll. By being yoked together, even in an unconnected state, the individual epiphanic units of this draft are not self-sufficient events, but rather blocks for a narrative-in-process.
A revision of the epiphanic praxis is interesting in that the epiphany is, according to Joyces definition, itself a form of re-vision. As you recall, the epiphany is explicitly defined only in Stephen Hero : By an epiphany he meant a sudden spiritual manifestation, whether in the vulgarity of speech or of gesture or in a memorable phase of the mind itself. He believed that it was for the man of letters to record these epiphanies with extreme care, seeing that they themselves are the most delicate and evanescent of moments (SH: 211). In an epiphany the soul or whatness of an object leaps to us from the vestment of its appearance (SH: 213). The epiphany thus has a two-fold aspect: on the one hand it is an experience of a sudden spiritual manifestation out of a relatively quotidian or mundane event, and on the other hand it is the artistic reproduction of that experience. The epiphany is thus not just the experience but the written account of that experience. Concomitantly, the epiphany is what defines the artist: the artist is the person who is able to record these spiritual manifestations with appropriate sensitivity. An epiphany is only an epiphany if it is recorded, it is the artistic après-coup of experience, the artists revision of experience.
Joyce himself revised this definition of the artists rôle in A Portrait of the Artist as a Young Man. There, Stephens argument elides the key-word epiphany and, instead replaces it with the more redoubtably Thomistic term claritas . According to Stephens argument, claritas, in a very general way, is the synthesis of sensible and intellectual perception as a transcendent revelation. Claritas is only to the extent that it can be communicated by the appropriately sensitive artistic soul: The image, it is clear, must be set between the mind or senses of the artist himself and the mind or senses of others (P213). This then would be a very traditional Hegelian æsthetic in that the work of art stands as an intersubjective object. But, the important point here is that claritas is a function of writing, and thus, inevitably, a function of rewriting and revising, even as the word epiphany is revised out of existence in the text of A Portrait of the Artist as a Young Man.
On the level of style, the major change between the epiphanies and Joyces later use of an epiphanic form (or something like an epiphanic form) in Giacomo Joyce and Ulysses involves a dispersal of sense through ambiguous collocation and construction. In short, the manifestation of claritas is ambiguated in its communication, and with the NLI Proteus we can now clearly see that this tactic of ambiguation is something that evolved and changed across various drafts.
Like the epiphanies, each of the units on the Proteus draft is focused. One case in point is the fifth unit, a meditation on Gods will and the Arian heresy: The soul created even He dare not destroy for the lex eterna that overshadows His will. Is that then the divine substance wherein Father and Son are consubstantial? Arius could tell me, mocking his own answer. Luckless heresiarch! In a Greek watercloset he breathed his last. Dead and dirty. The ways of the Creator are not our ways.  Stephen here, unsurprisingly, is thinking about Aquinas, specifically Aquinas postulate that the soul is immortal.  According to Aquinas formulation of the lex eterna, this postulate is inviolable and unalterable since divine will and eternal law are identical.  By definition, God cannot unwill the soul he has created. But Stephen, unlike God, is not always bound by Aquinas orthodoxy. Although Aquinas makes an analogy between, on the one hand, God and his creation and, on the other hand, an artist and his artefacts,  such an analogy is imperfect because the artist can always revise. An artist is unbound or uncoupled from the lex eterna since, as Stephen says here, the ways of the creator are not our ways. This line is absent in all subsequent versions of this passage in Proteus but it appears in Nestor as Deasys prelude to his grandiose claim that All human history moves towards one great goal, the manifestation of God (U: 2.380-81). As there are no extant early Nestor drafts, it is uncertain if this line here is an echo of Deasy or if Joyce discarded the line from Proteus and then inserted it into Nestor. There are other echoes of Deasy in this draft, but they are also problematic.  In any case, Stephens use of this line differs considerably from Deasys. Instead of expressing faith in divine reason and purpose, Stephen proposes that unlike God, the artists will is labile. The figure of Arius thus stands as a kind of artist, even as he suffered an ignominious death from diarrha. 
In the subsequent draft, Joyce preceded this unit with the ninth unit from the NLI draft, which describes the Frauenzimmer (although they are not called such on that draft). This unit ends with the line Mother, womb of sin.  On V.A.3, Joyce links these two units together through a discussion of the material matrix in which a human is created: Wombed in darkness I was too, made not begotten by them.  In this way, elements from one unit (in this case human birth) combine with and influence another. In subsequent drafts, the units revise each other.
Some of the changes between the NLI and Buffalo drafts are ironic. In the NLI draft of the handmaid of the moon passage, we read the earliest extant draft of a portion of Stephens vampire poem: a mans lips to her kiss, which was then revised on the same document to read mouth to mouths kiss. [ 12] On the subsequent draft, where this passage is greatly elaborated, Stephens comment about his composition of this line parallels Joyces own revisions from the preceding draft: Mouth to her kiss. No, you must have two of em. Glue em well together: Mouth to her mouths kiss.  It is as if Joyce, through Stephen, is describing his own path of revision.
Another comment Stephen makes about his artistic inclination can also be construed as Joyce describing his own process of revision. This one directly involves the epiphanies: Remember your epiphanies written on green oval leaves, deeply deep, copies to be sent if you died to all the great libraries of the world, including Alexandria (U: 3.14143). This passage first appears on the Buffalo Proteus draft, where initially Joyce wrote near libraries before parodically inflating Stephens pretension by amending this to read great libraries.  Stephen here disparages the epiphanic form he had practiced, a form that Joyce had used in the early drafting of this very episode, as we can see in the NLI draft. It is as if the epiphanic form is like the godlike artist in Stephens theory of creation, refined out of existence, indifferent, paring his fingernails (P: 215). The epiphany is revised out of existence out of the text it engenders, only then to be mocked. The artist is bound by no lex eterna .
A more intriguing revision involves Stephens metaphoric equation of language and the shifting beach-sand, which is inspired by the recollection of Louis Veuillots derisive description of Théophile Gautiers unwieldy prose as un coche ensablé .  The first version, on the NLI draft, reads: Heavy on this sand is all language which tide and wind have silted up.  On the Buffalo draft this is changed somewhat: The heavy sands are language that tide and wind have silted here.  On the Rosenbach manuscript, Joyce elided the preposition that, which improves the overall flow of the sentence at the expense of some syntactic clarity.  The basic sense is essentially the same in both versions. Stephen equates the shifting sands with language in that both are continually changing and evolving through external forces. Sands shift by the motion of wind and tide and language changes through evolving patterns of usage. In the Paradiso, Adam uses a different figurative image to explain languages mutability to the pilgrim: the usage of mortals [ie language] is like a leaf / on a branch, which goes and another comes.  Stephens line reflects the science of this episode as indicated on the Linati schema, philology, the study of the Protean nature of language.
Although the two versions have a congruous general sense, the specific articulation on each draft is subtly, but tellingly, different. In the first version, language is heavy on the sand; if you will, a metaphorical duvet. In distinction, in the second version, language is explicitly equated with the shifting sands: The heavy sands are language. Furthermore, the sense of the word silted is not necessarily identical between these two versions. In the first, it works in the sense of accumulating and being deposited by the wind and tide. In the second version, it works in the more figurative sense of flowing and passing away. One could say then that the sense of the word silted has itself silted between these two versions; one sense flows in as another dissipates across the palimpsest of drafts. Furthermore, the metaphorical equation of sand and language has also silted in that language becomes, between the two drafts, identical with sand rather than with its blanket. Stephens description of silting language is thus an apt metaphor for the linguistic changes made between the drafts of a work in progress. Between drafts, a new text comes that silts up and over the language of the preceding, receding draft.
The fragmentary or epiphanic disposition of the NLI draft is also an example of silted language. The final form of the episode bears little or no trace of this preliminary configuration; the epiphanic form has been silted over. And yet, now that this draft has resurfaced, certain traits of this revised, or revising, form of the epiphany can be made clear. For example, in reworking the Paris epiphany on the NLI draft, Joyce is no longer treating the epiphany as simply the representation of some particular and peculiar event within the world ie a phenomenological event but rather as a linguistic event, an event of an already-written, ie already-drafted, text that is being revised. In other words, and with other words, the epiphany is silted. If Joyce originally construed the epiphany as a mode of artistic re-vision, then the NLI draft suggests that the horizon of re-vision is entirely linguistic. This is, of course, exactly what is performed during any compositional revision; it is as if one draft serves as the epiphany for the next. Each new draft supplementing the one before and standing as another potential epiphany for some forthcoming subsequent draft. In other words, revision shows that the act of æsthetic apprehension and representation as described by Stephen is not immediate and ontologically self-sufficient. Protean, it evolves, it changes. It is dynamic. It silts over. It is also, potentially, an endless sequence of textual evolution.
Walter Benjamin discusses how in Hölderlins translations of Sophocles meaning threatens to become lost in the bottomless depths of language.  Well, Joyces penchant for continual emendation, much to his printers consternation, threatens a similar bottomless depth of revision. Revision is always a theoretically endless task. Indeed, were it not for the deadline of Joyces fortieth birthday, he might never have finished Ulysses and by finished I simply mean stopped.