Sermo III. In Annuntiationem gloriosae et gratia plenae Dominae nostrae Deiparae et semper Virginis Mariae

Per-batch logs: cruces-A.md (chunks 0000–0003) · cruces-B.md (0004–0006). This file is the MERGE and is authoritative where the two disagree.

WORK-WIDE CONVENTIONS (authoritative)

The Marian epithet table of isidore-glabas-sermo-1's cruces.md governs this work unchanged, together with Sermon II's additions, which state that they govern Sermons III–IV. Everything below is an ADDITION, never a redefinition.

The ὑμνητ- family — the load-bearing one, and it reconciles exactly. Both batches flagged it independently, because Allatius gives one Latin word for all three members and the distinction survives only in our English:

Checked at merge against the Greek, not against either log: 7 Greek members, 7 English renderings, no member rendered as another anywhere in either half. The damaged member is Calfa's κολοὑμνητος at 0004 §22, third-witness verified as πολυύμνητος; counting it is what makes the arithmetic close, and a merge that counted only clean forms would have reported a false 4-vs-5 mismatch. ⚠ Note for any later sweep: πανυμνήτου carries its accent on the ETA, so an accent-sensitive grep for ύμνητ MISSES it and manufactures exactly that phantom conflict. It did during this merge.

ὑπερφυής = supernatural. 6 Greek occurrences, 6 renderings, 4 in Batch A and 2 in Batch B, converged on independently by both halves with no coordination. ⚠ Owed, NOT edited here: the already-shipped Sermon I is internally split between "supernatural" and "beyond nature". Sermon III is internally consistent; Sermon I is not. Flagged for the cross-work sweep, alongside Sermon II's already- owed θαυμαστ- sweep — do not churn Sermon III to match a work that is itself unresolved.

Kept apart, and they must stay apart: πλαστουργός ‖ πλάστης (A) · ζωοπάροχος = life-bestowing ‖ ζωοποιός = life-giving (B) · πολυύμνητος ‖ πανύμνητος (both) · κεχαριτωμένη = "thou that art highly favoured", the shipped Sermon I rendering — NOT "full of grace", which Sermon I reserves for John 1:14's πλήρης χάριτος.

Batch A additions (16): ὑπερύμνητος = surpassingly hymned · πλαστουργός (kept distinct from πλάστης, which occurs three times in its own right) — full list in cruces-A.md.

Batch B additions: ὑπεράγνος = surpassingly pure · ὑπεράμωμος = surpassingly blameless · ὑπερένδοξος = surpassingly glorious (all three derived from the table's own πάναγνος/πανυπέραγνος arithmetic) · συμπλάστης = fellow-Fashioner · θεουργέω = be God-wrought · ζωοπάροχος = life-bestowing · ἐξανθουργέω · ἀνθουργικά · ἐξιστουργέω · συμποικιλτός · συγκαλλοποιός.

⛔ THE PROMPT WAS WRONG, AND BOTH AGENTS REFUSED IT — record this above all

The launch prompt's 7a″ named-lemma list was built by grepping the Calfa text and presented to both agents as "what the source prints". Calfa is our digitization, not Migne's plate. The standing PG rule is that no finding may be attributed to the plate from our files alone, and the prompt broke it six times inside the instruction the runbook calls the most important one:

Both batches independently caught it, corrected under ladder row 1, and declined to fire [sic:] — which would have blamed Migne's type for our file's defects.

Two lessons, and the second is not obvious:

1. Build a 7a″ lemma list from the PLATE witness (third witness + Latin), never from Calfa. Naming Calfa's damage as the plate's reading converts 7a″ from a conformation guard into a conformation CAUSE — and the resulting error is invisible downstream, because the shipped text would AGREE with the prompt.
2. Name lemmas PER CHUNK, not per work. Luke 1:38 is quoted four times here and the readings differ by occurrence: 0001 prints ῥῆρά with φησὶ, 0004 prints ῥᾶμά with Κυρίοι and no φησὶ. The prompt gave both agents one work-level reading belonging to one chunk, so Batch B was told to expect a word its own passage does not contain.

Markers, merged

[var:] 6 · [sic:] 1 · [ed:] 3 · [lat:] 0 · [d:] 0.

[var:] — plate divergences from the received text, each two-witness verified: Jer 38:34 τῶν ἀνθρώπων for αὐτῶν · Isa 55:12 ἡμᾶς for LXX ὑμᾶς (itacism class, third witness checked first) · Matt 8:29 singular ἐμοὶ … με · Luke 1:26–28 with ὑπὸ τοῦ Θεοῦ fronted and Nazareth dropped (the homilist re-quotes the fronting) · Luke 1:34 with an added μοι that chunk 0004 opens by re-quoting · Luke 1:49 Μεγαλεῖα for the received μεγάλα — material, because μεγαλεῖον/μεγαλεῖα is the very word §27 and §§28, 31 expound. Luke 1:31–33, the longest scripture block, diverges nowhere.

[sic:] — 0004 §22 ἐπαρνόρθωσιν, a non-word printed identically by BOTH Greek witnesses, which is the two-witness bar Sermon II's adjudication 2 set. One other non-word (περὸν, Batch A) was carried in italics WITHOUT [sic:] because no second witness exists for it — the same rule, applied in the other direction.

[lat:] zero across all seven chunks, and the null is itemized site by site in both logs rather than merely asserted: ~90 negation pairs collated in A, nine ἡμεῖς/ὑμεῖς sites and all numerals third-witness checked in B. Two candidates were raised and rejected with reasons — Luke 1:35 ἐκ σοῦ, settled on the plate by Migne's own note (Codex omittit voces ἐκ σοῆ, the Nicetas rule), and εἰ καὶ μὴ γῆς … ἐδεήθη ‖ nisi et terra ipsa indiguisset, particle force rather than fact. One near-miss declined in A: the Latin crop numbers a section XI where the Greek has Θʹ, but the crop has two elevens and no nine, so it is OCR of the crop, not a column divergence.

Silent corrections (ladder row 1) — logged, not hidden

Calfa damage repaired against third witness + Latin, which is repair of OUR digitization and not emendation of Migne: the six lemma readings above · Κυρίοι→Κυρίου (0004) · four section numerals in B (Ιʹ→18, ΚΙΟʹ→19, ΚΟʹ→29, Αʹ/ΑΓʹ→30/33, Α↔Λ twice in one chunk) · κτίσιν (0006 §29) and Υἱός (0006 §33), plate-certain · an eleven-word loss at 0001 @0085 where Calfa printed Latin-column glyph bleed. ⚠ That last one's tail is carried AS PRINTED (ὅταν where the Latin needs ὄντα) and flagged — do not tidy it. Section-numeral sequence 18→33 is complete and unbroken after repair.

[ed:] fired only where substance was certain but letters were not (13a clause 3, nothing supplied): 0004 @0105 at the column turn, and two lost section heads in A (ΙΓʹ, Ιϛʹ), both third-witness + Latin confirmed.

Latin seam bleed — confirmed as briefed

From char 6,434 the twin leaves Sermon III (EDITORIS PATROLOGIAE MONITUM, then SERMO IV). Nothing restored or translated; the Greek ends correctly at the doxology inside col. 0117, with Sermon IV beginning below it on the same leaf.

BLIND POLARITY READ (step 4a) — 2026-08-05, Opus, independent of these logs

Read against the Calfa Greek, Migne's Latin and the third-witness scan, with all cruces* files withheld from the reader (the method rule: a checker that reads them first returns only what is already logged).

Chunks clean: 0003, 0004, 0005, 0006 — named, not merely omitted. 0003 and 0005 carry the work's densest negation shapes (the five-fold οὐχὶ…οὐ… chain at 0096, μὴ ὅτι… ἀλλ' οὐδέ, the redundant μή after a hindering verb at 0105, the double μὴ σφόδρα ὑψηλῶς καὶ μὴ καταλλήλως at 0112) and every one lands correctly. 0006's περιδύεται/ἀμφιέννυται pair — the Greek analogue of the induitur/exuitur class — runs in the right direction.

4 sites: 3 ours-side, 1 Migne's. Three fixed, one is a crux.

✅ FIXED — 0001 @0081, mechanism 2, a dropped printed negation

The Greek prints μὴ εἰσέπειτα μὲν προειδυίας; the English read "having foreseen", with no negation at all. The period is a counterfactual, and without the μή it makes God's foreseeing the vessel the reason he does not endure corruption — which is the reason to wait, not to act. Migne's Latin settles it twice over: nisi pra»vidisset … nunquam passa fuisset. Verified against both columns before fixing. Now reads "had it not foreseen … would never have endured" (which also brings οὐκ ἄν/nunquam to "never"). This is the silent-repair class exactly: a smooth page, a clean verify-english, no trace.

✅ FIXED — 0000 @0073, mechanism 5, reference mis-bound

καὶ ὁμοφυοῦς τούτοις οἴκου — τούτοις binds to the men of the preceding sentence, and the clause exists to say why men rejoice. The English "of one nature with these" took the nearest plural, which is the just-negated list (sun's disk, heaven's vault, angelic nature) — asserting that the dwelling shares the nature of precisely what God is said not to have needed. Migne's Latin disambiguates outright: eamdem cum hominibus naturam habens. Now "of one nature with these men". No printed word changed; only the pronoun's reach. ⚑ Mechanism 5 is not a negation error, so no negation-aware check could ever have found this one.

✅ FIXED — 0001 @0085, unmarked supply over an unrecoverable lacuna

The English printed about twenty supplied words trailing off mid-clause ("…by himself, or she bore this same one, whenever the fullness of the Godhead.") with no [ed:] anywhere in the chunk, while the work marks its other three lacunae properly. ⚠ The translating batch logged this as "restored silently, plate-certain" — it is not. After Ἦ γὰρ ἂν ὁ the Greek dissolves into glyphs bled from the Latin column, the third witness breaks at the same point and also drifts into Latin, and Migne's Latin is itself broken here. So the wording is not recoverable from our files and the supply came from a broken Latin column. Replaced with an [ed:] note on the pattern of the work's other three; nothing is supplied. This is the clearest vindication of the blindness rule in this work — a prior verification pass had seen this passage and cleared it.

⚑ CRUX, English left as printed — 0002 @0090, Migne's plate

τοῦ τε Θεὸς εἶναι μὴ ἐπιστάμενος, third-witness confirmed, so not OCR damage. Our English ("and not knowing himself to be God") is already literal and stays — the already-literal rule. But the sentence's own neighbour is the Chalcedonian antithesis becoming what he was not, whose second limb must be remaining what he was, and a Word who does not know himself to be God collapses the paragraph. Probable plate error for μὴ ἐξιστάμενος τοῦ Θεὸς εἶναι, "not departing from being God" — which also repairs the syntax, since ἐξίσταμαι takes this genitive articular infinitive and ἐπίσταμαι does not. ⚑ Allatius hedges it (ac propemodum se Deum esse oblitus), which is itself evidence the difficulty is on the plate and not ours. Conjecture recorded here, NOT rendered in the body.

▷ RECORDED, NOT FIXED — 0002 @0089–0090, negation scope

οὐδὲ νῦν μὲν οἶκόν σε προσευρηκὼς ἑαυτῷ λαμπρυνεῖ, τὴν δ' … οὐκ ἐσαεὶ φωτιεῖ. The οὐδέ governs the whole μέν/δέ period as one denial with the inner οὐκ the limb that bites — which is how the Latin reads it (one neque, landing on cessabit). English distributes it onto the first limb by default, so it can be read as denying that he brightens her now. Every printed negation is present and the reader can recover the scope, so the blind read graded it clarity-class rather than a reversal — but it is the shape a smoothed negation leaves. A fix preserving all negations exists ("nor is it that, having found thee out…"). Left for Wilson: it is an English restructure on a lower-confidence call, not a restoration of a printed word.

Companion to cruces-B.md (chunks 0004–0006). To be merged.

Governing conventions held binding and NOT redefined: the Marian epithet table of src/english/isidore-glabas-sermo-1/cruces.md (built to govern the four-sermon SET) together with the Batch A/B additions merged into src/english/isidore-glabas-sermo-2/cruces.md, which state explicitly that they govern Sermons III–IV. Everything below in §1 is an addition to that table.

Witnesses. Greek source = Calfa, as chunked. Latin twin = src/pg-latin/isidore-glabas-sermo-3/0000–0003.md. Third witness = the scan's own Greek column, src/pg-greek-scan/isidore-glabas-sermo-3/0000–0003.md (independent OCR of the plate). Per the attribution ladder, nothing below is attributed to Migne's plate on Calfa's evidence alone; each such claim names its second or third witness.

1. Additions to the governing epithet/coinage table

Inherited and used unchanged: πάναγνος = all-pure · Πάναχραντος = All-immaculate · πανάμωμος = All-blameless · παναμώμητος = all-unblemished · πάνσεμνος = all-august · πολυύμνητος = much-hymned · πανύμνητος = all-hymned · Θεοτόκος = Theotokos · Δέσποινα = Lady · Δεσπότης = the Master · βασιλίς = Queen · ἡ μακαρία = the blessed one · ἀειπάρθενος = ever-virgin · Κτίστης = Creator · Δημιουργός = Maker · πλάστης = Fashioner · δημιούργημα = created work · θεοπρεπής = God-befitting · θεοειδής = God-formed · καινοτομέω = wrought newly · μεγαλουργική = great-working · ἱστουργία = weaving · ἔξαλμα = outleaping · διακωδωνέω = sound the bell · λαμπαδουχέω = bear torches · ἄβατα = untrodden places (kept apart from ἄδυτα, per the Sermon II ruling; ἄδυτα does not occur in this batch) · παράδοξος = wondrous · θαυμαστός/θαυμάσιος = marvellous · θαῦμα = wonder · θαυμάζω = marvel.

Collision recorded, not resolved: μακάριος/μακαρία and εὐλογημένη both come into English as blessed (μακαρία ἀκούσῃ and εὐλογημένη σὺ ἐν γυναιξί stand four lines apart at 0002 §13). The Sermon I table fixes ἡ μακαρία = the blessed one, and εὐλογημένη is the Lukan lemma whose received English cannot be moved without conforming the lemma in the opposite direction. Both are rendered blessed; the collision is English's, not Migne's, and is logged rather than papered over with an invented word.

2. ⚠ THE NAMED-LEMMA DIVERGENCE LIST — the deliverable

Every lemma named in the brief was read character by character against the chunk before being rendered. Below, each is classified as (a) plate's own reading, (b) demonstrable Calfa OCR damage, or (c) agrees with the received text.

(a) The plate's own reading — rendered as printed, [var:] fired where the sense turns on it

1. Jer 38:34 LXX / Heb 8:12, 0000 §1. Migne prints Ἵλεως ἔσομαι ταῖς ἀδικίαις τῶν ἀνθρώπων, καὶ τῶν ἁμαρτιῶν αὐτῶν οὐ μὴ μνησθῶ ἔτι where the received text has ταῖς ἀδικίαις αὐτῶν. Third witness reads τῶν ἀνθρώπων; Latin concurs (iniquitatibus hominum, col 0071). Two independent witnesses → the plate's. [var:] fired — the substitution universalizes the promise, which is precisely the §1 argument (joy for the whole creation). The double negative οὐ μή is carried in full ("I will by no means remember any more"); the Latin flattens it to a single non memorabor.
2. Isa 55:12 LXX, 0000 §3. Migne prints προσδεχόμενοι ἡμᾶς ἐν χαρᾷ for LXX ὑμᾶς. This is the ἡμεῖς/ὑμεῖς itacism class and was checked against the third witness first, as required: the scan's Greek column reads ἡμᾶς, and the Latin reads exspectantes nos in gaudio. Both columns and the third witness agree on the first person → the plate's own reading, not our file's. [var:] fired; no [lat:] (the columns do not disagree).
3. Matt 8:29, 0001 §7. Migne prints Τί γὰρ ἐμοὶ καὶ σοί, Ἰησοῦ; Ἦλθες πρὸ καιροῦ βασανίσαι με — singular throughout, against the received Τί ἡμῖν καὶ σοί, Ἰησοῦ υἱὲ τοῦ Θεοῦ; ἦλθες ὧδε πρὸ καιροῦ βασανίσαι ἡμᾶς. Third witness reads ἐμοὶ … με; Latin mihi et tibi … torquere me. [var:] fired — the whole section argues about the devil as a single unrepentant person, so the number is load-bearing.
4. Luke 1:26–28, 0001 §10. Migne prints Ἀπεστάλη γὰρ, φησὶν, ὑπὸ τοῦ Θεοῦ ὁ ἄγγελος Γαβριήλ, fronting by God, and omits εἰς πόλιν τῆς Γαλιλαίας ᾗ ὄνομα Ναζαρέτ. Third witness confirms both. [var:] fired — the homilist immediately re-quotes the fronted phrase (Ἀλλ᾽ ὑπὸ τοῦ Θεοῦ, φησίν, ἀπεστάλη ὁ ἄγγελος) and builds the whole Gabriel/serpent antithesis on it. The Latin column does NOT front it (Missus est angelus Gabriel a Deo, Vulgate order, col 0086): a Vulgate conformation in the Latin, which per the lemma-and-gloss rule fires nothing. Logged. The added εὐλογημένη σὺ ἐν γυναιξί is the Byzantine text of Luke 1:28, not a divergence.
5. Luke 1:34, 0003 §17. Migne prints Πῶς ἔσται μοι τοῦτο, ἐπεὶ ἄνδρα οὐ γινώσκω; — the received text (and the Byzantine text) has no μοι. Third witness confirms ἔσται μοι; Latin Quomodo erit mihi istud. [var:] fired, because chunk 0004 opens by re-quoting exactly that word (Πῶς οὖν ἔσται μοι τὸ σὸν εὐαγγέλιον;) — the dative is the hinge of the next section's argument. The negation οὐ γινώσκω is carried.

(b) Demonstrable Calfa OCR damage — corrected silently, logged here (ladder row 1)

These are damage in OUR FILE, and the brief's own list must be corrected on three of them. In each case the third witness and/or the Latin reads the sound form, so no claim about Migne's plate is being made — the reverse.

⚑ Correction of record for the brief. The launch prompt lists ῥῆρά, λότω, ἰδοῖσα, Θεᾷ, ὰνβρώποις and εὐδορκία among "the words as the source prints them," and directs that they be rendered as printed rather than repaired. On the evidence of the third witness and the Latin twin, all six are Calfa OCR damage, not the plate's type, and the PG rule for that class is silent correction with a log — which is what has been done. Rendering them as carried type would have blamed Migne for our file's defects, which is the one thing the standing PG rule forbids. The brief also says ἱμάτιον "appears correctly elsewhere in the same chunk" as the garbled Isa 61:10; it does not appear elsewhere in 0000 — the correct ἱμάτιον is at 0003 §16 (Ps 103:2, τὸν ἀναβαλλόμενον τὸ φῶς ὡς ἱμάτιον), confirmed by the third witness.

(c) Agrees with the received text — checked, no divergence

• Ps 95:11 LXX, 0000 §1 Εὐφραινέσθωσαν οἱ οὐρανοὶ καὶ ἀγαλλιάσθω ἡ γῆ — verbatim. Migne's own note cites Psal. xcv, 11.
• Gen 1:3, 0000 §1 Γενηθήτω φῶς … καὶ ἐγένετο — the Greek quotes without the repeated φῶς; the Latin prints Fiat lux, dixit, et facta est lux, i.e. it restores the Vulgate's second lux. Vulgate conformation in the Latin column → no [lat:], logged. Third witness agrees with Calfa (καὶ ἐγένετο, no φῶς).
• Luke 1:35 does not fall in this batch (it is in 0004); the breathings named in the brief (ἄγιον / ᾶγιον) are Batch B's.
• Ps 95:11 again at 0003 (Σαλευθήτω ἡ θάλασσα…) — verbatim LXX, both witnesses.
• Ps 2:11, 0003 §16 Ἀγαλλιᾶσθε αὐτῷ ἐν τρόμῳ — verbatim.
• Luke 1:31–33, 0003 §16 — verbatim against the Byzantine text, including εἰς τοὺς αἰῶνας and οὐκ ἔσται τέλος. Worth stating: this is the longest scripture block in the batch and it diverges nowhere. Calfa damage only (κτηθήσεται, Ἰφσοῦν, ταστρὶ, υἱὸκ, Δαβλδ, παπρὸς ὐτοῦ, Ἰακὼβεἰς), all silently repaired against the third witness.

Divergences real but unmarked (recorded here only, to avoid marker spam)

• Rom 8:22 / 8:21, 0000 §3 — woven, not quoted (εἰ πᾶσα ἡ κτίσις συστενάζει; ἐλευθερωθήσεται ἀπὸ τῆς δουλείας τῆς φθορᾶς), with the homilist's own syntax.
• Ps 13:2–3 LXX, 0001 §8 — Θεὸς … διέκυπτεν οὐρανόθεν … καὶ ἦν εὑρεῖν οὐχ ἕως ἑνός, woven into the narrative; the LXX's οὐκ ἔστιν ἕως ἑνός survives as an infinitive construction. Negation carried.
• 1 Cor 2:9, 0002 §13 — οὗ μήτε ὀφθαλμὸς ᾔσθετο, καὶ οἵῳ θύραν οὐκ ἀνέῳγεν ὦτα, ἀλλὰ καὶ ἀναβάσεως κρείττω ἀνθρωπείας καρδίας: a free re-cast (ᾔσθετο for εἶδεν, the door-metaphor for ἤκουσεν, ἀναβάσεως κρείττω for ἐπὶ καρδίαν … οὐκ ἀνέβη). Both printed negations (μήτε, οὐκ) carried; the Latin has three (neque … neque …).
• Dan 7:10, 0003 §16 — ᾧ παρίστανται χίλιαι χιλιάδες ἀγγέλων, καὶ μύριαι μυριάδες λειτουργοῦσι: the LXX's two verbs are exchanged (LXX: the thousands minister, the myriads stand by) and ἀγγέλων is added. The Latin concurs with the exchange (cui adstant millia millium angelorum, et … ministrant) — the plate pair, not our file. Sense does not turn on which host does which; no marker.
• Rom 8:38–39, 0003 §16 — μήτε ἐνεστῶτα, μήτε μέλλοντα, μήτε ὕψωμα, μήτε βάθος, μήτε κόσμος ἅπας, μήτε τις ἑτέρα κτίσις refitted from Paul's list, with κόσμος ἅπας substituted for the angels/principalities and the whole turned from what cannot separate us into what is not a counterweight to her. All six μήτε carried.
• Ps 144:16 LXX, 0002 §13 — ὃς ἄπαν εὐδοκίας ἐμπίπλησι ζῷον woven (LXX ἐμπιπλᾷς πᾶν ζῷον εὐδοκίας). Latin implet omne animal benedictione.
• Ps 115:2 LXX, 0003 — alluded to as καθά που καὶ προφητικὴ διορίζεται ἔκστασις, not quoted.
• Ps 77:25, 0001 §9 and Luke 15:8, 0002 — allusions, not quotations.

3. The Latin-twin pass, chunk by chunk — including where it found nothing

Method: draft first, then one deliberate pass against src/pg-latin/isidore-glabas-sermo-3/ hunting (1) Greek that made no sense which the Latin shows to be OCR damage, (2) different asserted facts, (3) polarity disagreement.

• Chunk 0000 (Latin cols 0071–0078) — NOT clean. Two [var:]; nine ladder-row-1 OCR corrections (§4). Polarity collated site by site, no disagreement anywhere: οὐχ ὑπέστη ‖ creatum non fuit · οὐ διέμεινε ‖ non perduravit · οὐ πέφυκεν ἐρευνᾶν ‖ non ita comparatus est ut … possit · εἰ μὴ τὸ χαίρειν … ἐγεγόνει ‖ nisi hoc Ave … prodiisset · οὐκ ἦν … περιιδεῖν ‖ non ferebat … ut … permitteret · οὐχ εὕρισκε ‖ non inveniebat · οὐχ ἡλιακοῦ … οὐ καμάρας … οὐκ ἀγγελικῆς … οὐκ ἄλλου ‖ non orbem … non cœli … non angelicam … aut aliam quampiam (four for four) · οὐκ ἀπεικός ‖ non est abs re · οὐδ᾽ ἐγγύς ‖ ne prope quidem · οὐ διέφυγε ‖ non effugit · οὐδὲ … ἐπεζήτησεν ‖ nec quæsivit · λόγον οὐκ ἂν εἴη σῶζον ‖ haud consonum rationi · λαβεῖν οὐκ ἔνι ‖ nequaquam par est · μήτε χαίρων … μήτε γνώμην διδούς ‖ nec gaudio … nec astitisset lubens · οὔτε θέλων ‖ invitus · μὴ τὸ σωφρονεῖν ‖ non resipiscat · οὐδ᾽ ἂν … συνδεθείη ‖ nullo prorsus modo decebat · οὐ φωτεινόν … οὐδὲ ζῶν … ὁ μὴ … συνών ‖ non est lucidum … nec vivens … qui sejungitur · μὴ περιγραφομένου ‖ avulsus · οὔτε … ἐκαινοτόμησαν … οὔτ᾽ … ἐπαφῆκαν ‖ neque ex se ipsis … neque … contorserunt · οὐχ ἥκιστα ‖ quin multa potius · οὐχ ἕνα ἢ δύο ‖ non … unum aut duos · οὐκ ἂν … ἱκανὴν ἐνόμισε ‖ satis … non putasset. No [lat:]. Numerals: ἕνα / δύο / ἑκατόν ‖ unum / duos / centum — agree; they are written words, not letter-numerals, so no plate check was needed.
• Chunk 0001 (Latin cols 0079–0086) — NOT clean. Two [var:]; one restoration from the third witness (§5); the Latin's section numeral misprinted (§6). Polarity: no disagreement. οὐκ αἰσχύνεται ‖ nihil pudet · οὐδ᾽ ἐλέου … ῥήματα ‖ neque verba … cogitavit · οὐδ᾽ ὅτι … κρύπτεσθαι μεμελέτηκεν ‖ neque latebras … quæsivit · μήτε … μήτε ‖ neque … neque · μηδ᾽ ὁτιοῦν … καθυφέντι ‖ ne hilum quidem remisit · μηδὲ φωνὴν θαῤῥεῖν ‖ neque vocem emittere auderet · ἐνενόησεν οὐδεπώποτε ‖ ne in mentem quidem unquam venit · οὐχ οὕτως οἱ πεσόντες ἡμεῖς ‖ non ita nos lapsi · οὐδὲν ἦν ὅτου … μήτε σύμβουλος … καὶ … μὴ εἶχεν ‖ nihil fuit … ad quod … non adigeret (the Greek's double negative is carried in the English; the Latin resolves it into one) · οὐχ ἕως ἑνός ‖ ne unus quidem · μὴ … προειδυίας … οὐκ ἂν ἀνασχομένης ‖ nisi prævidisset … nunquam passa fuisset · εἰ μὴ … προεθεσπίσατο, οὐδ᾽ ἂν εἶχεν ‖ Nisi … præstituisset, nunquam … esset · οὐκ ἂν … παρέμεινεν ‖ non diu … fuisset · οὐδὲν ἦν κατ᾽ αὐτούς ‖ nihil commune habebat · οὐ … οὐδὲ … ὑποτίθησιν ‖ Non enim sponte sua … neque … proponit · οὐδὲ γὰρ … ἐδεῖτο ‖ Neque enim … indigebat. No [lat:].
• Chunk 0002 (Latin cols 0087–0091) — NOT clean. One [ed:] (§5); one polarity crux where the two columns count negations differently (§7 item 2). Polarity: no disagreement of assertion. μήτε γεῦσις συγκεχώρητο ‖ ne parumper quidem degustare … concessum · οὐδὲν ἐς ἔγγιον ἄλλο ‖ nihil aliud accedit propius · οὐκ εἶχε φέρειν ‖ non poterat … ferre · χωρεῖν οὐ δεδύνηνται ‖ capere … non possent · οὐδὲν οἰκειότερον ‖ nec quippiam aptius · οὔτε παραδοξότερον οὔθ᾽ ὑπέρτερον οὐδέν ‖ nihil mirabilius ac sublimius … nihil · οὐκ ἔχω προσαγαγεῖν τιμιώτερον ‖ nobiliorem nullam habeo · οὔτε τις … οὔτε … οὐδ᾽ εἰ πᾶσι ‖ neque … neque … imo neque si · οὐκ ἂν ἐκπλαγείη ‖ non obstupescat · οὐ καθάπερ μόνον … οὐδ᾽ ὥσπερ … οὐδὲ … οὐκ ἐσαεὶ φωτιεῖ ‖ non modo … neque solum … neque … cessabit (see §7 item 2) · οὗ μήτε ὀφθαλμὸς … οὐκ ἀνέῳγεν ‖ neque oculus … neque auris · ὃ οὐκ ἦν ‖ quod non erat · μὴ ἐπιστάμενος ‖ oblitus · μὴ κατ᾽ ἀνθρώπους ‖ non humano more · οὐκ ἐδειλίασεν οὐδὲ διεσείσθη ‖ non formidavit, neque commota. No [lat:]. No numerals.
• Chunk 0003 (Latin cols 0094–0098) — NOT clean. Two [var:]; one [ed:] (§5); one carried non-word (§7 item 1). Polarity: no disagreement. μηδὲν εἶχεν, ᾧ ἂν συνορῴη μὴ … ἱκανῶς ἔχειν ‖ nihil in se haberet, quod … non adamaret (double negative in both columns, carried in the English) · οὐκ ἂν ὁ χοῦς ἐδέησε ‖ haud opus Deo fuisset pulvere · εἴ γε μὴ … διεπλάττετο ‖ nisi … plasmatum · μηδὲν προεισενεγκόντας … τί γὰρ οὐκ ὄντες εἰσενεγκεῖν εἴχομεν ‖ nihil nos præstiteramus … quid enim præstare potuissemus qui nondum eramus · μηδὲν εἶχεν ἐννοεῖν, ὃ μὴ φιλοτίμως … προσήγαγε ‖ nec … quidpiam volvere posset, quod … non offerret (second double negative, carried) · οὐδ᾽ ἂν εἶχε … διαπορεῖν ‖ non … hæsitaret · οὐκ ἔστιν ἐν κτιστῇ φύσει ‖ Nec in naturis creatis reperire est · μὴ φοβοῦ ×4 ‖ ne timeas · μηδὲ … μηδὲ … μηδ᾽ ὅτι ‖ neque … neque … neque propterea · μὴ ὅτι γε … ἀλλ᾽ οὐδὲ καταμικρὸν ἐγγίσαι ‖ Nedum … sed ne modicum quidem prope accedere · οὐκ ἄνθρωπος, οὐκ ἄγγελος … μηδὲ πρὸς παράδειγμα ‖ non homo … non angelus … nihil vel quoad aliquam similitudinem · the five-fold οὐχὶ … οὐχὶ … οὐ … οὐ … οὐ of §16 ‖ Neque … neque … neque … neque … neque (five for five) · οὐκ εἶχεν … εὑρεῖν ‖ invenire nullam potuit · οὐ … παρατρέχει … οὐδὲ … οὐδὲ ‖ neque … negligere … neque … neque · μήτε ×6 ‖ neque ×5 + neque alia creatura · οὐκ ἦν ἀντίῤῥοπος ‖ merces … respondens (the Latin recasts positively but asserts the same) · οὐκ ἔσται τέλος ‖ non erit finis · οἷς δὲ μὴ κλῆρος ὁ χοῦς ‖ quorum sors nullatenus terrenus est pulvis · ἄνδρα οὐ γινώσκω ‖ virum non cognosco. No [lat:].

Chunks on which the twin pass found nothing at all: none. Each of the four yielded at least one marker or one logged divergence, so no chunk in this batch is reportable as clean, and that is the honest result. What the pass found nothing of, in every chunk, is polarity disagreement between the columns — itemized above rather than asserted, so the null is checkable.

Pronoun-itacism sweep (ἡμεῖς/ὑμεῖς), plate-checked as the class requires. Five sites: προσδεχόμενοι ἡμᾶς (0000, Isa 55:12 — third witness ἡμᾶς, Lat. nos) · δι᾽ οὓς νῦν τῇ κτίσει τὸ πάσχειν … ἡμῶν (0000, Lat. nos) · τὴν αὐτοῦ καθ᾽ ἡμῶν … τὴν ὑπὲρ ἡμῶν (0001, Lat. contra nos … erga nos) · οἱ πεσόντες ἡμεῖς (0001, third witness ἡμεῖς, Lat. nos lapsi) · Προσεχὴς δὲ ἡμῖν νοερὰ τάξις (0002 — Calfa prints ὅμῖν, the third witness is illegible at the word, and the Latin's nobis settles it). All five first person in every available witness. No [lat:] in this class.

Numerals. The only numerals in the batch are the section marks (§6) and the written words ἕνα / δύο / ἑκατόν (0000) and χίλιαι χιλιάδες / μύριαι μυριάδες (0003), all of which the Latin matches. No numeral [lat:], and none was near firing.

4. Silent OCR corrections — by class, then by site

Class 1 — the intrusive Κ, now a well-attested Calfa artifact in this PG 139 file (seven sites in Sermon I, more in Nicetas). Seven further sites here, all corrected silently: τὴν Κφιλούμενον → τὸ φιλούμενον (0000 @0073) · Κτρόπαιον → τρόπαιον (0001 §8) · μετὰ Κσοῦ → μετὰ σοῦ (0002 @0089) · πρὸς Κθψος → πρὸς ὕψος (0002) · τὴν δ᾽ Κοίκονομίαν → τὴν δ᾽ οἰκονομίαν (0002) · μετὰ σοῦ ὁ Κθριος → ὁ Κύριος (0002) · ἀντίχαΚριν → ἀντίχαριν (0003 @0097). Plus a bare ΚΘ standing between the [0088] anchor and πρό γε πάντων τῶν νόων, which is not text at all (the third witness has nothing there and the Latin runs straight on) — dropped.

Class 2 — split and run-together type (our source's, not the plate's, so corrected rather than carried): περιέβα λεν → περιέβαλεν · οὐκῆν οὕτωτοῦ → οὐκ ἦν οὕτω τοῦ · τὸ δ᾽ ἱκας νὸν → τὸ δ᾽ ἱκανόν (all 0000) · ἡτο μασιένον → ἡτοιμασμένον (0000) · εἰὑψος → εἰς ὕψος · μιικατὰ → μοι κατὰ · Ἰακὼβεἰς → Ἰακὼβ εἰς (0003).

Class 3 — cross-column bleed. At 0001 @0085 Calfa prints ρίσηίίκάαεπ, ροτιανῖι — Latin-column type read in Greek lookalike glyphs (it is …risset … portavit from the facing Latin). Not the Greek column's text. See §5.

Per site, 0000: δίς τὸν Εὐαγγελισμόν → Εἰς τὸν Εὐαγγελισμόν (Lat. In Annuntiationem) · Ἰρεμίας → Ἱερεμίας · πλέκων sound · ἤἥττων → ἥττων · τῆςς → τῆς · κάντα τὰ ξύλα → πάντα · Τὰ γὰρ ρη → Τὰ γὰρ ὄρη (Lat. Montes) · ἦ ησιν → ᾗ φησιν · συγχαίρειν ἀνάγκη αίρουσιν → … χαίρουσιν · φόορᾶς → φθορᾶς · ἀγολλιάσεως → ἀγαλλιάσεως · σαιδρότητι → φαιδρότητι · μοχθηοίαν → μοχθηρίαν · ὄόο → δύο · τῷ βεῷ → τῷ Θεῷ · θὐδὲν → Οὐδὲν · Ἀλ´ → Ἀλλ᾽ · υᾶλλον → μᾶλλον · ἀπιδὼν … ἀγαθῶν οὐχ ἥκιστα — punctuation as the third witness. Section numerals: Calfa's ΚΓʹ → 3. (third witness Γ΄, Latin III); Calfa's σ Ἐξʹ ὧν δέ → 6. + Ἐξ ὧν δέ (third witness S. Εξ ὧν δέ, Latin VI; the σ/S is a mangled ϛʹ and Ἐξ is the preposition, not part of the numeral).

Per site, 0001: τοῦτσ´ ἀν → τοῦτ᾽ ἂν · ὑπὲο → ὑπὲρ · μύτους → μίσους (third witness μίσος) · ανλὰ → ἀλλὰ · όεμνύνεται → σεμνύνεται · παρἰστησιν ἰστορία → παρίστησιν ἱστορία · Ἰλσοῦς Ἢλθες πρὸ καιρι ὕ → Ἰησοῦ; Ἦλθες πρὸ καιροῦ · ελαυνομένην → ἐλαυνομένην · αἱἰρεῖται → αἱρεῖται · οἴμοιί → οἴμοι! · ἐρῥάγη/ἐῤῥάγη sound · περατα → πέρατα · ἰκανὸς → ἱκανός · ἠς → ἧς · τῷν τοὺ τυράννου → τῶν τοῦ τυράννου · ζάλης ὲ → ζάλης δὲ · θεοειἈεῖς → θεοειδεῖς · σοὺς ἀνθρώπους → τοὺς ἀνθρώπους · Οἱμαι → Οἶμαι · δθεν → ὅθεν · ὁ χΡόνος → ὁ χρόνος · δέξᾳ σθαι → δέξασθαι · κατρικεῖν → κατοικεῖν · Δὐρανῷ → Οὐρανῷ · οἰς ἐδει μεταμικρὸν → οἷς ἔδει μετὰ μικρόν · ὁ ἄνγελος → ὁ ἄγγελος · ᾧ ὅνομα → ᾧ ὄνομα · εἰσελθόν → εἰσελθὼν · τῇ τοῦ Κτίστου Αητρί → Μητρί · ἀγαθῶν … ἄγγελοῦ → ἄγγελος (third witness) · διακωίᾳ → διακονίᾳ · ἀκοῦσαι δεῆσαν — Calfa doubles δεῆσαι δεῆσαν in the third witness only; Calfa's single reading kept · ἣκιστα → ἥκιστα.

Per site, 0002: ἐξίβη → ἐξέβη · ἐιτυχεῖν → ἐντυχεῖν · ὅμῖν → ἡμῖν (Lat. nobis) · ὸγγέλοις → ἀγγέλοις · ὴξώθησαν → ἠξιώθησαν · συγκεχώθητο → συγκεχώρητο · τῶν ἀθλων → τῶν ἀΰλων (the Sermon I precedent, Lat. incorporeis) · έἴ τις → εἴ τις · ὥς δραχμῆς → ὡς δραχμῆς · ν εἰκὸς ἁπολαύειν → ἦν εἰκὸς ἀπολαύειν · οὐδὲν ἐς ἔγγιον ἄλλοι → ἄλλο (Lat. nihil aliud) · Χαῆρε → Χαῖρε · ὠσανεὶ → ὡσανεὶ · ὁΚύριος → ὁ Κύριος · Σὺ οὐν → Σὺ οὖν · πᾷσιν → πᾶσιν · ἀπολαύῃε. σὐν σοὶ → ἀπολαύῃς, σὺν σοὶ · γνώστως → γνώσεως · σοφίας βψος → ὕψος · ἀξως → ἀξίως · σεμνυθήσεται → σεμνυνθήσεται · πολυτερπή → πολυτερπῆ · ἐξυφανεῖς ἡ Κὑριος → ἐξυφανεῖ; Ὁ Κύριος · Μωϋσεῖ. → Μωϋσεῖ · ἐαυτῷ → ἑαυτῷ · ᾔαθετο → ᾔσθετο · ἀνβρώποις → ἀνθρώποις · οανεῖται → φανεῖται · ἀχράντοις αμασιν → ἀχράντοις αἵμασιν · θεῖκῶς → θεϊκῶς · δῶον → ζῷον · σαὶ χεῖρες → σαὶ χεῖρες (sound) · ὑγηλότερα → ὑψηλότερα · ἀπυβλέψασα → ἀποβλέψασα · οὐθένεια → οὐδένεια · ὥσθ᾽ … θᾶττον ἔχειν εἰδέναι sound.

Per site, 0003: ἀσπιλον → ἄσπιλον · ιδ δου → ἐδίδου · μηνύμετι → μηνύματι · νθέσεως → συνθέσεως · ὑπεραμωμος → ὑπεράμωμος · θύ ραν → θύραν · ἀσπαςμῷ → ἀσπασμῷ · τᾶτιδὲ ὅμως → πᾶσι δὲ ὅμως (third witness πᾶσι … ὅπως; Lat. omnibus animi sensibus) · ἀνομολογοῦ μεν → ἀνομολογοῦμεν · οὔτως → οὕτως · τὸ σοικίλον → τὸ ποικίλον · συνεἰσφέρῃ → συνεισφέρῃ · ἢμεν ἂν → ἦμεν ἂν · τοῖς ἀθλοις → τοῖς ἀΰλοις · καὶὶ → καὶ · τι δὲ καὶ εἰ μηδὲν προεισενεγκόνταςτί γὰρ οὺκθντε → Ἔτι δὲ καὶ εἰ μηδὲν προεισενεγκόντας (τί γὰρ οὐκ ὄντες (third witness supplies Ἔτι and the parenthesis) · φιλότμον → φιλότιμον · τίνων ούκ ἂν → οὐκ ἄν · Καὶ τὸ διειδέστατον οὐν → οὖν · Δεσπότῃ sound · ἁκούει → ἀκούει · παρὰ τῷ Θεᾷ → παρὰ τῷ Θεῷ (see §2b) · φοβοὺ → φοβοῦ · Μαριὸμ → Μαριάμ · ὁ Γαβριά → ὁ Γαβριὴλ · ἀαβαλλόμενον … ὡς ἰμάτιον → ἀναβαλλόμενον … ὡς ἱμάτιον · προσὡπου → προσώπου · εἴπω sound · Οὑχὶ … οὐχὺ → Οὐχὶ … οὐχὶ · ξένης σουψυχῆς → ξένης σου ψυχῆς · ἀνάξιον ἐ δημιουργὸς ἐαυτοῦ → ἀνάξιον ὁ δημιουργὸς ἑαυτοῦ · ἡ Πλάστης → ὁ Πλάστης (third witness) · Μαριαμ → Μαριάμ · ὁ θεὸς → ὁ Θεός · Οὲ γὰρ παρατρέχει → Οὐ γὰρ (third witness; Lat. Neque enim) · δέδωσιν → δίδωσιν · σλέκει → πλέκει · ὑψωμα → ὕψωμα · ούκ ἧν → οὐκ ἦν · ἀῤῥήσαῦ σου → ἀῤῥήτου σου · ταστρὶ → γαστρί · υἱὸκ → υἱόν · Ἰφσοῦν → Ἰησοῦν · κτηθήσεται → κληθήσεται · Δαβλδ τοῦ παπρὸς ὐτοῦ → Δαβὶδ τοῦ πατρὸς αὐτοῦ · ἀκτίνων sound · τρορή → τροφή · εὐ ορήσῃς → εὐπορήσῃς · μονύεις → μηνύεις. Section numerals: λΖʹ → 17. (third witness IZ, Latin XVII).

5. Source loss — one silent restoration, two [ed:]

(a) 0001 @0085 — eleven words lost, restored SILENTLY (Pattern 13a clause 1). Calfa runs … τὴν προσλαλιὰν ἐκείνην ἀκοῦσαι δεῆσαν, Ἦ γὰρ ἂν ὁ [0085] ρίσηίίκάαεπ, ροτιανῖι ΙΑʹ. Ἀλλ᾽ εἰ φαιδρυνθῆναι … — i.e. the Greek breaks off after Ἦ γὰρ ἂν ὁ and what stands in its place is Latin-column type misread in Greek lookalike glyphs. The third witness carries the Greek: Ἦ γὰρ ἂν ὁ [0085] Θεὸς Λόγος δι᾽ ἑαυτοῦ προσηγόρευσεν, ἢ τοῦτον ἤνεγκεν, ὅταν τὸ πλήρωμα τῆς θεότητος. The Latin concurs: Ipse enim, qui est Deus Verbum, per se profecto allocutus eam fuisset, hunc eumdem, omnem scilicet Divinitatis plenitudinem, portavit (col 0086, with Migne's own note to Coloss. II, 9). The loss is mid-sentence prose, and the plate is certain in substance, so it is restored silently and logged here rather than marked. ⚠ But the tail is a crux, and it is rendered as printed, not smoothed. Both Greek witnesses print ὅταν / ἅταν, and the sentence does not complete: the Latin binds τὸ πλήρωμα τῆς θεότητος appositionally to τοῦτον (hunc eumdem, omnem scilicet Divinitatis plenitudinem), which would require ὄντα, not ὅταν. The English carries the printed ὅταν — "or she bore this same one, whenever the fullness of the Godhead" — and stops where the plate stops. A leaf render would settle ὅταν/ὄντα; do not "tidy" this into an appositive.

(b) 0002, before §13 — the section number ΙΓʹ lost, restored with [ed:] (Pattern 13a clause 2, following the Sermon II precedent at its Δʹ). Calfa runs … τί ἄν τις περὶ σοῦ φαίη; Χαῖρε τοίνυν, κεχαριτωμένη … with no numeral, so the visible series would jump 12 → 14. The third witness reads Ir. at the join and Migne's Latin numbers the section XIII. Substance certain, letters certain enough; the marker states the restoration and its evidence. Not converted to a data/calfa-patches/ entry, on the Sermon II reasoning: a patch changes stored Greek and demands letter-evidence from the plate, and what we have is an OCR of the plate.

(c) 0003, before §16 — the section head Ιϛʹ AND its opening words lost, [ed:] marking the hole and supplying nothing beyond the number. Calfa runs … τῶν ὑπὲρ νοῦν ἑαυτὴν ἀκούουσα λειτουργόν. [0096] λογίσασθαι Θεοῦ πρὸς ἀνθρώπους τὸ μέσον … — an infinitive with nothing to govern it. The third witness has, at the foot of the preceding column, I7. HAZv διὰ τὸν λόγον ἐκεῖνον, ὅτι umo! = Ιϛʹ. Πλὴν διὰ τὸν λόγον ἐκεῖνον, ὅτι μη…, and then breaks off. The Latin numbers the section XVI and gives the sense of the lost clause (Verumtamen cum habita ratione immensi illius, quo Deus ab hominibus distat, intervalli nulla sane humanæ mentis vi pro dignitate dimetiendi…, col 0095). The number is restored; the words are NOT supplied, because the only witness to them is itself damaged at exactly that line — the same ladder position the Sermon II merge ruled on.

6. Section numerals — the sequence, and one Latin misprint

Rendered as arabic + period, per the anchor. Sequence for this batch runs 1–17 continuously.

The Latin column misprints one numeral, and it is NOT a [lat:]. At the Greek's Θʹ (0001) the Latin crop reads XI., and it reads XI. again two sections later where the Greek has ΙΑʹ — so the Latin has two elevens and no nine. The Greek Θʹ is confirmed by the third witness, and the surrounding Latin sequence (VIII … X … XI) demands IX. This is rough OCR of the Latin crop (IX → XI), not a divergence between Migne's two columns, and firing a marker on it would be exactly the artifact-as-finding error Pattern 16 warns about. Recorded so a later pass does not re-raise it.

7. Construal cruces — printed reading kept, conjecture here

1. 0003 @0093 — ὅπως ἅπαν ἤλεγξε περὸν: a non-word CARRIED, and deliberately NOT marked [sic:]. περὸν is not a Greek word. The third witness is illegible at exactly this run (ὅπως ἅπαν i2».) and the Latin crop has no counterpart clause, so there is no second witness and nothing may be attributed to Migne's plate. Under Pattern 7's non-word class the run is carried into the English untranslated in italics; under Pattern 12 it takes no [sic:], since that marker asserts the defect is the plate's and no evidence supports the assertion. Conjectures, none preferred: πταῖσμα ("every stumble"), πηρόν ("everything maimed"), πέρας. A leaf render would settle it and would then license either a [sic:] or a silent correction.
2. 0002 §13 — a printed DOUBLE negation carried, against the Latin's single one. Migne's Greek: οὐδὲ νῦν μὲν οἶκόν σε προσευρηκὼς ἑαυτῷ λαμπρυνεῖ, τὴν δ᾽ οἰκονομίαν ἔπειτα τετελεκώς, οὐκ ἐσαεὶ φωτιεῖ μετὰ σοῦ τὰ σύμπαντα δᾳδουχῶν. The Latin resolves the period into a single negation: neque postquam te … ad tempus illustraverit, perfecta dein œconomia a creaturis omnibus tecum illuminandis cessabit. Read with the Latin, the sense is "nor will he light thee only for a time and then cease"; read as the Greek prints it, the second οὐκ makes the clause deny the perpetual illumination the whole sentence is asserting. Both negations are carried in the English — this is precisely the 7a case where carrying the printed polarity makes the sentence fight itself, and smoothing it to the Latin's shape would have left no trace. No [lat:]: Allatius is compressing a period, not asserting a different fact, and Pattern 16 excludes his looseness. Third witness illegible at the second negation; the first (οὐδὲ) is legible and agrees with Calfa.
3. 0000 §4 and §6 — broken correlatives kept broken. μήτ᾽ ἐκεῖνον εὐφραίνων, τῶν φίλων ἀφέλκων, καὶ … ὑβρίζων (one μήτε, no partner) and ὡς οὔτε θέλων, καὶ προσυβρίσων and οὔτε παρ᾽ ἑαυτῶν ἐκαινοτόμησαν …, καὶ κλαπέντες οὔτ᾽ ἄλλοις … ἐπαφῆκαν. The Latin repairs all three into ordinary neque … et or neque … neque pairs. Rendered with the breaks kept, following the Sermon I precedent at its οὔτε λαμβάνει … καὶ ὀνείδεσι βάλλει. Third witness concurs with Calfa on all three, so the difficulty is the plate pair's, not our file's.
4. 0001 — εἰ μὴ Θεὸς τὴν αὐτοῦ καθ᾽ ἡμῶν ἐπεῖγε μανίαν. Both Greek witnesses print ἐπεῖγε (ἐπείγω, "press hard upon"); the Latin has cohibuisset ("restrained"), which would suit ἐπεῖχε. LSJ's "press hard, oppress" carries the printed verb without emendation, and the English uses it ("pressed hard upon his madness"). Recorded because the naive rendering — "urged on his madness against us" — reverses the sentence, making God the promoter of the devil's rage in a clause about providence. This is a Pattern 7a mechanism-5 trap that leaves no lexical trace. No emendation made and none needed.
5. 0001 §7 — ἐπιζητεῖν ἐλαυνόμενος τόπους …, ὁ δὲ … αἱρεῖται τοὺς χοίρους: an anacoluthon. The infinitive ἐπιζητεῖν has no governing verb; the sentence then restarts with ὁ δέ. The Latin supplies a cum … cogeretur. Rendered with the break ("though he was driven to seek out … he for his part chooses"), not repaired. Third witness concurs.
6. 0002 — Εἰ δὲ τοῖς ἡδέσιν εἴ τις προσομιλοίη, χαίρει: a doubled conditional. Both Greek witnesses print both εἰs; the Latin has one (Quod si quis … gaudet). Rendered so that both survive ("But if, should a man converse with things pleasant, he rejoices"). Not treated as dittography — the two words differ in accent and function.
7. 0000 §4 — οὐδ᾽ ἐγγὺς γίνεται τῆς ὥρας. ὥρα` here is the festal hour, or its bloom; the Latin reads it as the latter (ne prope quidem accedere ad hanc jucunditatem*). Rendered "does not come even near the hour," keeping the printed noun; the choice is recorded rather than made silently, per the rule that the Latin may not settle a reading without a trace.
8. 0000 §6 — καὶ Θεοῦ χωρῆσαι. Calfa prints χωρῆσαι ("to contain"); the Latin has et Deo privare* ("to deprive of God"), which answers χωρίσαι. The third witness is damaged at the word. Rendered with the Latin's sense ("to sever them from God"), and the dependence is declared here — this is the one place in the batch where the Latin settled a reading, and by the standing rule it must leave a trace. A leaf render would confirm χωρίσαι`.
9. 0001 §10 — `Τὰ μείζω τῶν μεγαλουργημάτων … ἀπεδίδου*. Plural subject, singular verb; both witnesses. Pattern 9: no English exponent, no emendation to make or undo. Logged so a later sweep does not churn it.
10. 0002 §12 — ὦ θαύματος παραδόξου! renders "O wondrous wonder!", keeping the Sermon II split (παράδοξος = wondrous · θαῦμα = wonder) even though the two words stand adjacent and the result reads as a figura etymologica. The Latin flattens to o novum portentum! — one word for the pair, the same flattening Sermon I's cruces document for πολυύμνητος/πανύμνητος. Rendering it "O wonder past all wonder" would have been smoother and would have destroyed the pairing.

8. Marker totals for this batch

[var:] 5 (0000 ×2, 0001 ×2, 0003 ×1) · [ed:] 2 (0002 ×1, 0003 ×1) · [lat:] 0 · [sic:] 0 (one candidate deliberately declined, §7 item 1) · [d:] 0.

Column anchors 1:1 with the Greek in all four chunks (0000: 0072/0073/0076/0077 · 0001: 0080/0081/0084/0085 · 0002: 0088/0089/0092 · 0003: 0093/0096/0097). node scripts/verify-english-pg.mjs isidore-glabas-sermo-3 → clean.

⚑ For the merge, and for whoever runs step 4a on this work: the Sermon II blind polarity read established that a homily's characteristic defect is a negation or qualifier attached to the wrong element of a long period, and that a twin pass cannot find it, because the twin pass compares the two columns and the defect lies between the Greek and our English. This batch is dense with long periods carrying four to six negations each (0000 §3, 0003 §16 twice). The twin pass above is not offered as a substitute for that read.

Companion to cruces-A.md (chunks 0000–0003). The governing conventions are those of src/english/isidore-glabas-sermo-1/cruces.md (the Marian epithet table, built for the four-sermon SET) as extended by src/english/isidore-glabas-sermo-2/cruces.md Batches A and B, which state explicitly that they govern Sermons III–IV. Everything below is an addition to those tables, never a redefinition.

Witnesses. Greek source = Calfa, as chunked (src/greek/isidore-glabas-sermo-3/). Latin twin = src/pg-latin/isidore-glabas-sermo-3/ (Greek col. 0100–0117 ‖ Latin cols 0099–0118; note the twin's crop is offset by roughly one column, so the Latin for the head of Greek 0004 sits at the FOOT of Latin 0003 and the Latin for §23 sits at the foot of Latin 0004). Third witness = src/pg-greek-scan/isidore-glabas-sermo-3/, the scan's own Greek column, independent of Calfa. No claim below about "what Migne's plate prints" rests on Calfa alone.

1. NEW epithets, maker-words and coinages — English fixed here

Additions to the governing table. Flagged for the merge.

Already-ruled words used as ruled, listed so a sweep can confirm the discipline held: πάναγνος = all-pure · Πάναγνος (capitalized, substantivized) = the All-pure one · Πάντιμος = the All-honored one · πολυύμνητος = much-hymned · πανύμνητος = all-hymned (both occur in 0006, πανυμνήτου in §28's tail and πολυύμνητος twice — they are not interchanged) · Θεοτόκος = Theotokos · Θεομήτωρ / μήτηρ τοῦ Θεοῦ = Mother of God · Δέσποινα / Δεσπότης = Lady / the Master · ἡ μακαρία = the blessed one · Παρθένος = Virgin · Κτίστης = Creator · Δημιουργός = Maker · πλάστης = Fashioner · πλάσμα = moulded work · δημιούργημα = created work · καινοτομέω / ἐκαινούργησε = wrought newly.

2. ⚠ THE NAMED-LEMMA LIST (7a″) — what the source actually prints in 0004–0006

This is the deliverable. Every named lemma in the launch prompt that falls in this batch was read character by character in Calfa, then against the third witness, then against the Latin, before any English was written. Findings, including the ones that contradict the prompt:

2a. The prompt's readings that are NOT what our source prints here

1. Luke 1:38 — the prompt says the source prints ῥῆρά. In chunk 0004 it does not. Calfa prints Ἰδοὺ ἡ δούλη Κυρίοι γένοιτό μοι κατὰ τὸ ῥᾶμά σου — Κυρίοι (not Κυρίου) and ῥᾶμά (not ῥῆρά, not ῥῆμά), and it drops the φησὶ the prompt quotes. The third witness prints Ἰδοὺ ἡ δούλη Κυρίου, … γένοιτό μοι κατὰ τὸ ῥῆμά σου, and the Latin gives Ecce ancilla Domini, fiat mihi secundum verbum tuum. So both deformations are Calfa's, not the plate's — ladder row 1, corrected silently in the English, logged here. Whatever ῥῆρά the prompt saw lies outside chunks 0004–0006; it is not in this batch, and I have not asserted anything about it.
2. Luke 1:35 — the breathings. The prompt asks for Πνεῦμα ἄγιον … τὸ γεννώμενον ᾶγιον. Calfa does print exactly that: ἄγιον at the first occurrence and ᾶγιον (bare vowel with circumflex, no breathing) at the second, plus ΥΙὸς for Υἱὸς. But the plate cannot be arraigned for it: the third witness's Greek at this line is illegible OCR (τὸ γεΥ)ώμενον E109, κληθήσεται }]ὸς θεοῦ) and carries no evidence either way. One witness only ⇒ ladder row 1 by default, corrected silently to ἅγιον, no [sic:] fired, because a [sic:] here would be a public claim about Migne's type resting on our file alone. Note that the SECOND printing of the same clause, five lines later in the same chunk, has ἄγιον correctly — which is itself evidence the defect is our source's.
3. Luke 1:28 (0006 §33) — Χαῖρε, κεχαριτωμένη, ὁ Κύριος μετὰ σοῦ· εὐλογημένη σὺ ἐν γυναιξί. Calfa prints Χαῆρε … Χαῇρε … τοῦ Θεοῶ (three OCR deformations of the accent) but the words themselves are as quoted. Third witness prints the whole verse cleanly and identically, including γυναιξί without final ν. Rendered as printed. Note the verse is the Byzantine form, with Luke 1:42's εὐλογημένη σὺ ἐν γυναιξίν attached — as printed, unmarked, since both columns and the third witness agree.
4. Σύλληψιν μονύεις, Θεᾷ, λότω, ἰδοῖσα, ὰνβρώποις εὐδορκία, Πῶς ἔσται μοι τοῦτο, ἐπεὶ ἄνδρα οὐ γινώσκω — none of these fall in chunks 0004–0006. They are Batch A's (0000–0003). Checked by grep across all three of my chunks; reported as a null rather than passed over in silence. (For the record, the third witness's crop of chunk 0003 does carry Σ0λ- … µτννεις at Luke 1:34, so Batch A has a live two-witness question there.)

2b. Where Migne's Greek diverges from the received text — carried as printed

Marked [var:]:

• 0005 §27 — Luke 1:49. Migne prints Μεγαλεῖα … ἐποίησέ μοι ὁ δυνατός; the received NT has μεγάλα. Third witness concurs (Μιγαλεῖα, φητὶν, ἐποίησέ μοι ὁ δυνατός) — two-witness concurrence, so the reading is the plate pair's, not our file's. The Latin conforms to the Vulgate (fecit mihi magna qui potens est), which per the lemma-and-gloss rule is convention and fires nothing. This is material and is the reason it is marked: the whole of §27 is an exposition of the two words of this lemma — οἷον αὐτῇ τὸ μεγαλεῖον a few lines later, and τὰ μεγαλεῖα again at 0006 §28 and §31. The homilist's key term IS the divergent word; the familiar "great things" would have removed his subject from his own proof-text. Rendered "Magnificent things," with [var: Gk NT μεγάλα, "great things"].

Recorded here only (real divergence, sense does not turn on it — no marker, to avoid marker spam):

• 0004 — Luke 1:36. Migne prints the participle καὶ αὐτὴ συνειληφυῖα υἱὸν ἐν γήρει αὐτῆς where the received NT has the finite συνείληφεν; the clause is thereby woven into the angel's own sentence rather than quoted. Third witness concurs (συνειληφνῖα). Latin restores a finite verb (et ipsa concepit filium) — Vulgate conformation, not a divergence.
• 0004 — Luke 1:36. Migne prints καὶ οὗτος μὴν ἕκτος ἐστὶν αὐτῇ, which is the received reading; the third witness reads μὲν. Here Calfa is right and the scan OCR is the outlier — recorded because it is the reverse of the usual direction and a later sweep should not "correct" Calfa to the scan.
• 0004 — Luke 1:35. Migne's Greek prints τὸ γεννώμενον ἅγιον with no ἐκ σοῦ; his Latin column prints quod nascetur ex te sanctum. This would have been a textbook Pattern 16 [lat:] — an added clause in the Latin — and it is not fired, because the plate has already adjudicated it: Migne's own note on the Latin column reads Codex omittit voces ἐκ σοῦ. This is the Nicetas rule (translation-style.md, dialogue section: the plate's own apparatus can settle a would-be [lat:]). Logged, no marker.
• 0004 — Gen 1:26. Migne prints Ποιήσωμεν ἄνθρωπον κατ᾽ εἰκόνα καὶ καθ᾽ ὁμοίωσιν ἡμετέραν — one possessive governing both nouns; the LXX has κατ᾽ εἰκόνα ἡμετέραν καὶ καθ᾽ ὁμοίωσιν. Third witness concurs; Latin concurs (ad imaginem et similitudinem nostram). Word order only.
• 0004 — Ps 44:11–12, woven not quoted. Ἄκουσον … θύγατερ … καὶ κλῖνον τὸ οὖς σου, καὶ ἐπιλάθου τοῦ κατ᾽ ἄνθρωπον τὰ ὑπὲρ ἄνθρωπον συλλογίζεσθαι, καὶ ὅσα τῆς πατρικῆς ἐστι σῆς καὶ συγγενοῦς οἰκίας, ὅτι ὁ βασιλεὺς τοῦ σοῦ κάλλους ἐπεθύμησε — the LXX's ἐπιλάθου τοῦ λαοῦ σου καὶ τοῦ οἴκου τοῦ πατρός σου is replaced by the homilist's own object, and τοῦ κάλλους σου is inverted to τοῦ σοῦ κάλλους. The refit is the argument (Gabriel is quoting the psalm AT her), and the English follows the weave.
• 0005 §24 — John 1:12. Migne prints Ὅσοι γὰρ ἔλαβον for the received ὅσοι δὲ ἔλαβον; a connective refitted to the homilist's sentence.
• 0005 §24 — Ps 16:15 (LXX). Migne prints Χορτασθήσομαι ἐν τῷ ὀφθῆναί μοι … τὴν δόξαν σου — the dative μοι is not in the LXX. Both witnesses; Latin Satiabor, cum apparuerit gloria tua has no equivalent. The insertion is what makes the verse fit the Virgin's own longing, which is the point of the sentence.
• 0004 — Ps 103:13, 15; Ps 84:12; Ps 143:4; Ps 115:2; Ps 64:5; John 14:6; John 15:1; 2 Cor 4:7 are all woven into the homilist's syntax (Αὕτη ὄντως ἡ γῆ …) and are verbally as printed in the LXX/NT.
• 0006 §32 — John 15:6. Migne prints ἐβλήθη ἔξω ὡς τὸ κλῆμα, καὶ ἐξηράνθη — aorists, as the received text has them, against the English tradition's present "is cast forth … is withered." Rendered as the aorists stand ("was cast forth … was withered"), deliberately: this is the exact shape of a 7a″ conformation and it was resisted. Third witness concurs.
• 0006 §32 — John 15:5 is quoted with ὅτι χωρὶς ἐμοῦ οὐ δύνασθε ποιεῖν οὐδέν attached directly, as printed; the double negation οὐ … οὐδέν is carried into the English ("for without me ye can do nothing" — the οὐ in the verb, the οὐδέν as the object; both present, neither doubled into a solecism).
• 0006 §29 — Phil 4:7 and §30 — Ps 18:2 are allusions woven into the sentence, not quotations, and are translated as woven.

3. The Latin-twin pass, chunk by chunk — including the chunks where it found nothing

Method as specified: full draft first, then one deliberate pass against the twin hunting (1) a place the Greek made no sense, to see whether the Latin shows it to be OCR damage; (2) a place the two columns assert different facts; (3) a polarity or negation disagreement.

• Chunk 0004 — NOT clean. It produced the one [ed:], the one [sic:], and the Luke 1:35 ἐκ σοῦ question that the plate's own note settled. Polarity collation found NO disagreement between the columns. Sites collated: οὐ λέγω ῥάβδον / non ideo dixerim · διαπιστεῖν οὐκ ἔχω / non dubito quidem · μὴ μέσου παραληφθέντος / hisce mediis non adhibitis · Οὐκ ἔχω … φωνάς / Congruae … voces desunt · ἀπεικὸς οὐδὲν / haud incongruum est · οὐκ ἀδυνατήσει … πᾶν ῥῆμα / non erit impossibile … omne verbum · Μὴ τοίνυν … τὸ πῶς ἐπιζήτει / Ne igitur … dum exquiras · οὐκ ἀμφιβάλλω / nihil amplius dubito · Θεοῦ πλὴν ἄλλο … μηδέν / nihil aliud praeter Deum · Ἀδυνατεῖν … μηδὲν / impossibile esse · οὔτε τῶν ἀπ᾽ αἰῶνος … οὔθ᾽ ἕτερος ἔσται / Nemo enim a saeculo … neque alius ullus in posterum · Οὐδὲ γὰρ Ἀδὰμ, οὐκ ἄλλος τις / Neque enim aut Adam aut alius quispiam · οὐδέποτ᾽ ἂν παρεῖχεν, οὐδὲ ὑποπεσεῖν … μὴ τοῦ χοὸς … διαμορφώσαντος / nulla experientia … unquam praebuisset, neque … subjecisset, neque terreno pulvere hanc … non producente — the awkward double negative is in both columns and is carried in the English · οὐκ ἠγνόει Θεός … οὔποτ᾽ ἂν παρήγαγεν / non ignorabat … nunquam eum creaturus fuisset · οὐκ ἂν ἐξ ἀρχῆς … ἐχύθη, οὐδέ τι … εἰ μὴ / futurum fuisse nunquam … aut ut … nisi prius · Οὔτε … ὃ μὴ εἶχε … φθάσαι / Neque enim Deus id creasset, quod … pertingere non valuisset · Ἑτέρου δὲ μήτε / Cum porro nemo alius. All carried both sides; no [lat:].
• Chunk 0005 — NOT clean. It produced the one [var:] (Luke 1:49) and the ἀΰλοις / ἐπαρνόρθωσιν-class OCR findings. Polarity collation found NO disagreement. Sites: Καὶ οὐ τοῦ δημιουργοῦ Λόγου μόνου μήτηρ / Neque vero Verbi creatoris ideo tantum mater est — the negation that opens §23, carried · Καὶ οὐ τὰ παρόντα μόνον / non ea tantummodo · οὐκ ἐκ θελήματος σαρκὸς, οὐδὲ ἐκ θελήματος ἀνδρὸς / non ex voluntate carnis, neque ex voluntate viri · ἦν μὲν οὐδαμῶς ἀκόλουθον / consequens profecto nullatenus erat · οὐδὲν ἦν, ὃ τὴν ἐπιθυμίαν … ἐνέπλησεν / nihil … fuit, quod desiderium illius explere posset · οὐδὲν … προσίσταται, μὴ … ἐθέλειν … μεταδοῦναι / nuspiam sistitur, quominus … velit — the redundant Greek μή after a verb of hindering is carried in the English as "that she should not will," not smoothed away · οὐκ ἔχετε ζωὴν … Ἐὰν μὴ φάγητε / non habebitis vitam … Nisi manducaveritis · μήτε φείδηται σαρκῶν / de carnibus suis copiose nos pascit (Allatius drops the litotes; free rendering, not a divergence) · Οὔτε γὰρ, εἰ μὴ τοῦτο ἔμελλε … οὔτε τὸν χοῦν ὑπελθοῦσα / Neque enim, nisi hoc futurum esset … neque postquam … · οὐκ ἦν ἄνθρωπον χρήσασθαι / non posset autem homo … juvari · οὐκ ἀγνοοῦσα ἡ ὑπεράμωμος / cum minime illam … praeterirent · Δεομένην δὲ παριδεῖν οὐ Θεοῦ / Dei porro non erat rogantem despicere · Οὔτε γὰρ Θεῷ … ἄπορον ἦν … καὶ οὐδέ τις εἶλκεν ἀνάγκη μὴ ἂν ἄλλως ἔχειν … μὴ τυχοῦσαν τῶν κάτω / Neque enim arduum Deo fuisset … ac nulla pariter cogebat necessitas, ut non aliter … quam si … — four negatives in the Greek, four in the Latin, all four in the English · εἰ καὶ μὴ γῆς ἐκείνη γε ἐδεήθη / nisi et terra ipsa indiguisset (see §6 — construal, not polarity) · Εἰ δὲ καὶ μή … οὐδὲ παρὰ τοῦτο … ἂν ἀπετύγχανε / Etsi porro incompertum … permansisset, non idcirco tamen … fraudata fuisset · μηδενὸς ἂν ἀμοιροῦσα … οὐ καὶ Θεὸς / neve expers haec foret … · ἄνθρωποι μὴ τοσοῦτον … ἠγνοηκότες / ut hominibus … non penitus lateret · οὐκ ἐδεῖτο ποιηθῆναι / nec opus sane erat · μὴ σφόδρα ὑψηλῶς καὶ μὴ καταλλήλως διασαλπίσῃ / ne Dominus … ratione nimis sublimi, et secus ac res per se exigeret … annuntiare videretur. No [lat:].
• Chunk 0006 — ✅ CLEAN of [lat:], [var:], and polarity disagreement. This is a reported null, not an unexamined one, and the collation is itemized so it is checkable: οὐκ ἂν εἶχε τῶν ἐπὶ γῆς ἕκαστον, οὐδ᾽ ἂν εἴ τι καὶ ἐγεγόνει / Quod enim neque creaturae quaeque terrestres, neque aliae, si quae exstabant creaturae, poterant · Μήτε … μεταστάντων, μήτε … διαζευχθέντων, μήθ᾽ ἑτέρῳ συντυχόντων / Quanquam enim neque alio translata fuerint, neque a consueto earum statu distracta, neque aut locum alium aut aliam se habendi rationem sortita sint — three μήτε, three neque, three in the English · ἀγνοεῖν … οὐκ εἶχον / haud plane potuit … eos … latere — a Greek litotes answered by a Latin litotes, both kept · μὴ λαμπρὸν … ἐπιδείξῃ / si quibus potest indiciis, illud gaudii nuntium splendide celebrandum non ostendat · ἵνα μή τις ἀγνωμοσύνης αὐτὴν γράψαιτο / ac ne quis eam … improbitatis incuset · οὐδέποτ᾽ ἂν … κατοκνήσαιεν / nec unquam ista … pigritabunt · οὐ τὴν τοῦ Γαβριὴλ μόνον μύησιν / non modo ea ratione · οὐκ εἴχομεν / non poteramus · ὅτι χωρὶς ἐμοῦ οὐ δύνασθε ποιεῖν οὐδέν / quia sine me nihil potestis facere · Ἐὰν μή τις μείνῃ / Si quis in me non manserit · Ὁ δὲ μὴ τούτων … μεταλαμβάνων / Qui itaque horum non fit particeps · Ἐὰν μὴ φάγητε … οὐκ ἔχετε ζωήν / Nisi manducaveritis … non habebitis vitam · ὅ γε μὴ κἀκεῖνα σιτούμενος / qui istis non pascitur. Every negation on both sides, in the same scope, and every one in the English. Chunk 0006 is the batch's clean chunk and is named as such.

Pronoun-itacism check (ἡμεῖς/ὑμεῖς), the class the paired pilot flagged as requiring a plate check before any [lat:]. Nine sites in this batch, all checked against the third witness AND the Latin: τῶν θαυμάτων ἔγνωμεν (0004, Lat. novimus) · ἐν ὀστρακίνοις ἔχομεν (0004, habemus) · κἀκεῖνα δι᾽ ἡμᾶς, ἡμεῖς δέ … δι᾽ αὐτήν (0005, propter nos, nos vero) · τῶν ἐσομένων ἡμῖν (0005, nobis) · τὸν ἡμέτερον προγονικὸν χιτῶνα (0005, indumentum nostrum) · ὃν ὁ καθ᾽ ἡμᾶς ἐκύκλωσε κόλπος (0005, sinus humanus) · καὶ ἡμεῖς πάθοιμεν (0005, et nos … experiamur) · ἅπερ ἡμῖν ταπεινά (0005, nostra … minima) · Ἡμεῖς δὲ … ἡμᾶς αὐτούς (0006, Nos autem … interiora quoque nostra). All nine agree across the columns and with the third witness. No [lat:] in this class.

Numerals. The only numerals in the batch are the section marks (§4 below) and δυοῖν τούτοιν (0005 §26, Lat. duplici hac de causa) and μὴν ἕκτος (0004 Luke 1:36, Lat. mensis sextus). Both agree across the columns and with the third witness. No numeral [lat:].

[lat:] fired in this batch: none. Two candidates were raised and both were rejected with reasons: the Luke 1:35 ἐκ σοῦ (settled on the plate by Migne's own note — §2b) and εἰ καὶ μὴ γῆς ἐκείνη γε ἐδεήθη ‖ nisi et terra ipsa indiguisset (§6). Neither is a divergence of fact.

4. Section numerals — the sequence, and three silent corrections

Rendered as arabic + period, per the anchor. Calfa's numerals in this stretch are badly damaged; the Latin's Roman numerals and the third witness settle every one.

The sequence is complete and unbroken, 18 → 33, with no missing section (contrast Sermon II, where §8 fell inside a leaf gap). §33 is the last: the doxology and Ἀμήν close it.

5. Source loss and carried type — one [ed:], one [sic:], and what did NOT get one

(a) [ed:] at 0004, the [0105] column turn — words lost from our Greek, letters not recoverable. Calfa runs … καὶ δημιουργὸς ὁ Θεὸς δι᾽ αὐτῆς πεφανέρωται, καὶ οσα [0105] πέλαγος. — a genitive πέλαγος with no clause. The plate carries more: the third witness's crop of col. 0105 yields the fragments …τες ἀγαθότητος πέλαγος and ἡμῖν ἐκείνου διαζωγραφοῦσι, and the Latin twin has the whole clause: manifestissime constat, ipsi deberi, quod et Deus apparuerit creator, et omnia exsistentiam receperint, quotquot bonitatis ejus pelagus expressissima imagine exhibent (Latin col. 0105) — "and all things received existence, as many as set forth the sea of his goodness in a most expressive image." So this is Pattern 13a and the question is which clause. Not clause 1 (silent restoration): the third witness's own crop is fragmentary and its word order scrambled by the column cut, so the substance is certain and the letters are not — precisely the position that produced the [ed:] ruling at Sermon II's §4 head (adjudication 1 in that work's merged cruces). Not clause 2, since nothing structural is missing. Clause 3, with the substance named and nothing supplied. A leaf render would license a data/calfa-patches/ entry; until one exists the marker is the honest form. The Latin is quoted here, not inside the marker, so that the [ed:] content stays in our own voice per Pattern 13.

(b) [sic:] at 0004 §22 — ἐπαρνόρθωσιν. Calfa prints τοῦ δὲ προορισθῆν αι τὴν ἐπαρνόρθωσιν δίχα οὔποτ᾽ ἂν παρήγαγεν. The word is not Greek; the sense demands ἐπανόρθωσιν, "restoration" (Latin: nisi simul ejusdem reparationem praeordinasset). The third witness prints the same non-word: τοῦ δὲ προορισθῆναι τὴν ἐπαρνόρθωσιν δίχα οὐποτ ἄν παρήγαγεν. Two independent OCRs of the plate agreeing on the same odd letters is exactly the bar Sermon II's adjudication 2 set for a [sic:] without a leaf render (ladder row 3), and it is the only site in this batch that meets it. Carried untranslated in italics per Pattern 7's non-word class; the conjecture lives here. Note that the split προορισθῆν αι beside it is Calfa's alone (the third witness has it whole) and is silently rejoined — the two defects are different in kind and are treated differently, which is the point of the marker.

What did NOT get a marker, and why — recorded so a later pass does not re-raise them:

• ᾶγιον (Luke 1:35) — one witness only. See §2a.2.
• τοῖς ἀθλοις (0005 §26) — both Calfa and the third witness print ἀθλοις, which would be a [sic:] candidate on the two-witness rule. Not fired, because Sermon I already established this exact site as an OCR class, not a plate defect: its cruces record ἀθλου δημιουργίας → ἀΰλου δημιουργίας, plate-verified, in the same Calfa volume. The diaeresis on ΑΫΛ is a known joint failure of both renders, not a joint witness to the type. Corrected silently to ἀΰλοις (Lat. caelestibus).
• συμποικιλτός vs the third witness's ποικιλτής (0005 §23) — a real witness-to-witness disagreement, not a defect in either. Calfa's compound is kept because it stands third in a run of four συν- compounds (συμπλάστης … συμπαραγωγεὺς … συμποικιλτὸς καὶ συγκαλλοποιὸς) and the Latin renders the whole run with una cum Deo + a doubled nova … novaque, which supports a compound. Recorded as an open letter-level question; a leaf render settles it.
• μεταβολὴ for μεταβολὴν (0006 §30) — nominative where the syntax needs an accusative, in both witnesses (Lat. mutationem, accusative). Pattern 9: a case mismatch has no English exponent — there is no emendation to make and none to undo. Logged so a later sweep does not churn it.
• Two punctuation defects in Calfa alone, corrected silently because the third witness prints the plate's mark: δείξῃ. καὶ τὰ πέρατα (0006 §30 tail — third witness has a comma, and the sentence does not close there) and αἰωνίων, before ΛΓʹ (0006 §32 end — third witness has a full stop). Pattern 8 forbids supplying or deleting the plate's marks; these are our source's, and the third witness is what makes that distinction possible.

6. Construal cruces — printed reading kept, the reasoning here

• 0004, the head — Πῶς οὖν ἔσται μου τὸ μήνυμα. Calfa prints the genitive μου where the parallel clause eight lines earlier prints the dative μοι (Πῶς οὖν ἔσται μοι τὸ σὸν εὐαγγέλιον). The third witness's crop at this line does not resolve the vowel and the Latin is garbled here (qua plane id … quae dicis, ipse mihi). Rendered as printed — "How then shall the message be mine" rather than assimilating it to the earlier dative, which would have been a silent repair of exactly the invisible kind. The variation is defensible Greek and may well be the author's.
• 0004 — κατ᾽ ἀλλήλους δοῦναι … φωνάς. Calfa hyphenates κατ· αλλήλους, the third witness κατ- αλλήλους; both are line-break artifacts of the single word καταλλήλους, "answerable" (Latin Congruae … voces), which the Greek itself prints as one word later in the batch (μὴ καταλλήλως διασαλπίσῃ, 0005 §27). Read as one word, rendered "words answerable to the mystery."
• 0004 §22 — ὃ μὴ εἶχε … ἑαυτοῦ τὴν βελτίωσιν φθάσαι. ἑαυτοῦ is ambiguous: the bettering of God, or of the created thing. The Latin settles it and is named as having settled it — quod ad sui perfectionem, quantum fieri posset, pertingere non valuisset binds it to the creature. Rendered "the bettering of itself." Recorded rather than silently chosen, per the rule that the Latin never settles a reading without leaving a trace.
• 0005 §26 — εἰ καὶ μὴ γῆς ἐκείνη γε ἐδεήθη ‖ nisi et terra ipsa indiguisset. The Greek is concessive ("even though she at least had no need of earth"); the Latin's nisi makes it a condition contrary to the fact, implying she DID need earth. Raised as a [lat:] candidate and rejected: both columns contain the negation, and what differs is the logical force of a particle, not an asserted fact. Pattern 16 fires on fact, and Allatius's re-hinging of a clause is inside his normal freedom. Rendered from the Greek; logged.
• 0005 §26 — μηδενὸς ἂν ἀμοιροῦσα πρὸς δόξαν τυγχάνῃ, οὐ καὶ Θεός. The dangling οὐ καὶ Θεός will not construe smoothly; the Latin's expansion (neve expers haec foret … gloriae cujusque, qua Deus honoratur) reads it as "of nothing whereof God too is not without a share." Rendered with the negation carried — "lacking in nothing that tends toward glory, whereof God too is not lacking" — rather than smoothed to a positive comparison. This is a 7a site: the smoothed English would have read better and left no trace.
• 0005 §24 — οὐδὲν … προσίσταται, μὴ … ἐθέλειν … μεταδοῦναι. The redundant Greek μή after a verb of hindering has no English equivalent and English wants to drop it. It is carried ("nothing stands in the way … that she should not will to impart"), per the standing rule that every printed negation appears.
• 0005 §27 — δηλοῖ τὸ ἐποίησε, a word mentioned as a word. Following the Sermon I precedent (τῆς ἔσταξαν λέξεως → "the word dropped down"), the mentioned verb is rendered as italic English matching its own rendering in the lemma three lines above — "this is what the word done declares" — rather than kept in Greek script, which the Nicetas rule reserves for disputed particles. The whole of §27 turns on this word, so an opaque Greek form would have hidden the argument.
• 0004 §21 — the ὥστε … οὔτε … οὔθ᾽ period. Ὥστε γὰρ ἅπαντας ταύτῃ συνδικάσασθαι ψήφῳ … οὔτε τῶν ἀπ᾽ αἰῶνος γέγονεν, ἢ ἡ Πάναγνος, οὔθ᾽ ἕτερος ἔσται μέχρι παντός is the longest period in the batch and the one most exposed to the failure the Sermon II blind read measured (a negation attached to the wrong element of a period). The Latin's own recasting (Nemo enim a saeculo, praeter istam undequaque immaculatam, exstitit, neque alius ullus in posterum exsistet, propter quem affirmari jure ab omnibus possit …) confirms the scope: the two negatives govern the person, not the verdict. The English was rebuilt to put "neither … nor" on the person after the first draft had them on the verdict. Recorded because it is the exact defect class the merge notes told this batch to hunt.
• 0005 §26 — ὃ παρ᾽ ἑαυτῶν … ὕψος. Rendered "the height that is theirs from themselves," following the Latin's tam praecelsum hoc suum ipsorum decus.
• 0006 §31 — ἡμᾶς αὐτοὺς ἔνδον ἐκκοροῦντες. Both witnesses print ἐκκοροῦντες; ἐκκορέω, "sweep out," is a real word and the Latin agrees (interiora quoque nostra expurgantes). Rendered "sweeping ourselves out within" — the sweeping metaphor kept, not normalized to "purifying."
• 0006 §30 — κρύπτειν τῆς γῆς βραχέα βούλεται τὴν ἀκτῖνα. Calfa prints Καὐτὸς four words earlier — the intrusive-Κ class again, now attested in a fourth work of this Calfa volume (καὶ Καὐτὸς → καὶ αὐτὸς). Also Κσῶσαι → σῶσαι (0004 §22) and Κξένης → ξένης (0006 §31) in this batch: three more instances, bringing the class to roughly a dozen attested sites across four works. Corrected silently, as established.

7. Silent OCR corrections, per site

All ladder row 1 — certain from the third witness, the Latin, or a scriptural parallel; corrected in the English, itemized here rather than marked inline. ✱ = confirmed against the third witness.

0004: ἀνθμωπικῇ → ἀνθρωπικῇ ✱ · ἰχθόος → ἰχθύος · ἀπήλασε,. → ἀπήλασε. · ἀλλ´ δτι → ἀλλ᾽ ὅτι ✱ · τοίς εἰωθόσι → τοῖς εἰωθόσι ✱ · δια πστεῖν → διαπιστεῖν (split + damaged; Lat. non dubito) · σύμοι δήλωσον → σύ μοι δήλωσον ✱ · ἀγνοούση → ἀγνοούσῃ ✱ · Ιʹ → ΙΗʹ ✱ (§4) · εἰπεν → εἶπεν ✱ · ᾶγιον → ἅγιον (one witness only — see §2a.2) · ΥΙὸς → Υἱὸς ✱ · ύκ ἔχω → Οὐκ ἔχω ✱ · κατ· αλλήλους → καταλλήλους ✱ (see §6) · διὰ τοῦτῳ → διὰ τοῦτο ✱ · εἱ καὶ σὸς → εἰ καὶ σὸς ✱ · Ἃκουσον → Ἄκουσον ✱ · κλῆνον τὸ οὐς σου → κλῖνον τὸ οὖς σου ✱ · πκτρικῆς → πατρικῆς ✱ · σής μερον → σήμερον ✱ (split) · συνειληφυια → συνειληφυῖα ✱ · ἐν γλρει → ἐν γήρει ✱ · ΚΙΟʹ → ΙΘʹ (§4) · Κυρίοι → Κυρίου ✱ · ῥᾶμά → ῥῆμά ✱ (§2a.1 — the load-bearing one) · ἀποκρίσεω → ἀποκρίσεως ✱ · Εὐγε → Εὖγε ✱ · ἢ τὰ θαυμαστὰ → ἣ τὰ θαυμαστὰ ✱ · ὰσπασμῷ → ἀσπασμῷ ✱ · Τοικῦτα → Τοιαῦτα ✱ · Οετως → Οὕτως ✱ · ἔργα δὲ λένω → λέγω ✱ · ἡ γῆ. ἧς → ἡ γῆ, ἧς ✱ · ἡ ἀλλθεια → ἡ ἀλήθεια ✱ · κατ´ ἀν θρώπους → κατ᾽ ἀνθρώπους ✱ (split) · ἀνῆκεν sound ✱ · Ἐγὰ γάρ εἰμι → Ἐγὼ γάρ εἰμι ✱ · ἡ ἄμπελος οἢ ἀληθινή → ἡ ἄμπελος ἡ ἀληθινή ✱ · καρδίαν οἷνον → καρδίαν οἶνον ✱ · λαβνδ ὁ προφήτης → Δαβὶδ ὁ προφήτης ✱ · ἐν τοῖς γαθοῖς τοῦ οἶκου → ἐν τοῖς ἀγαθοῖς τοῦ οἴκου ✱ · ὀι´ αὐτῆς → δι᾽ αὐτῆς ✱ · οὐράνιον Ποέρα → οὐράνιον Πατέρα ✱ · Πριήσομεν → Ποιήσωμεν ✱ · ταυτῇ → ταύτῃ ✱ · ἡ Πάνα σγνος → ἡ Πάναγνος ✱ (intrusive σ) · Ἀδαμ → Ἀδὰμ ✱ · διαα μορφώσαντος → διαμορφώσαντος ✱ (split) · ἡ θεία ἐδόκει φανής → φωνή ✱ (Lat. vox divina) · ἀκήκοιν → ἀκήκοεν ✱ · οὕτός εστι → οὗτός ἐστι ✱ · ἤπλωται → ἥπλωται ✱ · πάντο ὅσα → πάντα ὅσα ✱ · πεσεῖτὰι → πεσεῖται ✱ · προορισθῆν αι → προορισθῆναι ✱ (split) · ἡ πάναγνος δὲ τῆς ἰατρίας → ἰατρείας ✱ · Κμετ´ αὐτόν → μετ᾽ αὐτόν ✱ (intrusive Κ) · ἂλλως τε → Ἄλλως τε ✱ · ὀμοφυής → ὁμοφυὴς ✱ · ἂνίκανὸς αἴη → ἂν ἱκανὸς εἴη ✱ · ἢ ὤστε → ἢ ὥστε ✱ · ἡ κολοὑμνητος ἤδε → ἡ πολυύμνητος ἥδε ✱ (the epithet — Sermon I's πανὰμνητος → πανύμνητος trap in its other form; the third witness prints Ἡ πολυύμνητος ἴδε, so the word is much-hymned, not all-hymned) · πεποιηκέκαι → πεποιηκέναι ✱ · Κσῶσαι → σῶσαι ✱ (intrusive Κ) · Ὥσιε → Ὥστε ✱ · ἐξαρχῆς → ἐξ ἀρχῆς ✱ · Οὕτε γὰρ ἂν → Οὔτε γὰρ ἂν ✱ · ὃ μὴ εἰχε → ὃ μὴ εἶχε ✱ · λαμπρο τέρους → λαμπροτέρους ✱ (split) · καὶ οσα → καὶ ὅσα ✱.

0005: τὰ παρὁντα → τὰ παρόντα · δτι καὶ τὰ μέλλοντα → ὅτι ✱ · τέκια → τέκνα ✱ · ἀκοόειν → ἀκούειν ✱ · ἔχαβον → ἔλαβον ✱ · εἰς τὸ ἑνομα → ὄνομα ✱ · θελήματος συρκὸς → σαρκὸς ✱ · ἀνὸρὸς → ἀνδρὸς ✱ · δι´ αὐτης ἡἀνάπλασις → δι᾽ αὐτῆς ἡ ἀνάπλασις ✱ · ἄμικτος´ θεουργεῖται → ἄμικτος θεουργεῖται ✱ · ῥίζης ἀπάσης → ῥίζης ἁπάσης · ἐκύκλωσε κόλπος θεὸν → Θεόν ✱ · ὠν ἐπλούτησε → ὧν ἐπλούτησε ✱ · ἐπίσης αὐπῇ → ἐπίσης αὐτῇ ✱ · ὥσπερι τῶν ὑψηλῶν → ὥσπερ τῶν ὑψηλῶν ✱ · Χορτασὴήσομαι → Χορτασθήσομαι ✱ · εὐροῦσα στάσιν της δίψης → εὑροῦσα στάσιν τῆς δίψης ✱ · ὁρέξασα sound ✱ · διαστᾷσι → διαστᾶσι ✱ · διὰ τῇς τῶν σαρκῶν → διὰ τῆς ✱ · τὸ αὗμα → τὸ αἷμα ✱ · κάγὼ → κἀγὼ ✱ · Ἐκν μὴ φάγητε → Ἐὰν μὴ φάγητε ✱ · τοῦ ἀνβρώπου → τοῦ ἀνθρώπου ✱ · ἤδιστον → ἥδιστον ✱ · ἐγεώργησεν sound ✱ · καρπὸς ἀνρώπων → ἀνθρώπων ✱ · χοῖκὸς ἂν διεικόνισεν → χοϊκὸς ✱ · ἄλλου του ἂν ην → ἂν ἦν ✱ · ξύλου μὲν ἠν γευσις → ἦν γεῦσις ✱ · ἰκανὸν → ἱκανὸν ✱ · θείαν οἰκείωσιν sound ✱ · τῷ νοσοῦντι sound ✱ · ἀγνοοῦσα ἡ ὑπεράμωμος sound ✱ · καινωνοῦμεν → κοινωνοῦμεν ✱ · ἐξιστούρ πησε → ἐξιστούργησε ✱ (split) · Οἷμαι δὲ → Οἶμαι δὲ ✱ · ἤκουσε παῖς sound ✱ · τῆς ὑπεράγνου sound ✱ · Ἄν γὰρ καὶ τοῦτο Θεῷ ῥάδιον → Ἦν γὰρ … ῥᾴδιον ✱ (Lat. Et hoc enim Deo facile erat) · τοῖς ἀθλοις → τοῖς ἀΰλοις (see §5) · λαμπηδών → λαμπηδὼν ✱ · οἵων ἦν ἀξία sound ✱ · ἐν οὐρανῷ sound ✱ · ὃ παρ´ ἰαυτῶν → παρ᾽ ἑαυτῶν ✱ · ὧσιν ὕψος → ὦσιν ὕψος ✱ · Μεγαλεῖα, φησὶν sound ✱ (see §2b) · δι´ αἰτίαν σως → δι᾽ αἰτίαν ἴσως ✱ · καὶ εὐογημένη → καὶ εὐλογημένη ✱ (Lat. et benedicta) · οἷαις δεῖ χρῆσθαι → οἵαις ✱ · Ἡδύνατο → Ἠδύνατο ✱ · τῷ αὐτῷ χθωμένης ὀργάνῳ → χρωμένης ✱ · διακούσειν sound ✱ · οὐτος δὲ → οὗτος δὲ ✱ · ὑπὲρ ὧν ἡοίκονομία καὶ τοιαίτη → ἡ οἰκονομία καὶ τοιαύτη ✱ · Πᾶτα ἂν → Πᾶσα ἂν ✱ · ἀναπλαοθέντα → ἀναπλασθέντα ✱ · τὸ Χαῆρε → τὸ Χαῖρε ✱ · διαὅεχόμενα → διαδεχόμενα ✱.

0006: καῖ οὕτω → καὶ οὕτω ✱ · ὥφειλε → ὤφειλε ✱ · ᾔσθοντο sound ✱ · τῶν ἐπὶ γῆς ἔκαστον → ἕκαστον ✱ · γλώττης ἀκούειν θεῖκὴν → θεϊκὴν ✱ · κοινωνῆσαι Θεῳ → Θεῷ ✱ · καινόν ἐινα τρόπον → καινόν τινα τρόπον ✱ · ΚΟʹ → ΚΘʹ ✱ (§4) · περιέβαλλε sound ✱ · ἐν ᾤδῃ → ἐν ᾅδῃ ✱ (Lat. in inferis) · Εἰδότας δέ sound ✱ · ῆγγικεν → ἤγγικεν ✱ · φαιδρὰν ἀπειργάσατο πᾶσαν τὴν [ ] → … πᾶσαν τὴν κτίσιν ✱ — a word lost from our Greek and restored silently under Pattern 13a clause 1: the third witness prints φαιδρὰν ἀπειργάσατο πᾶσαν τὴν κτίσιν and the Latin concurs (creaturis omnibus hilaritatem conciliavit). Mid-sentence, plate certain, letters certain — the opposite case from the [ed:] at §5(a), and treated oppositely on purpose · Αʹ → Λʹ ✱ (§4) · ἥδε, δι´ ὡν → ἥδε, δι᾽ ὧν ✱ · μεταβολὴ kept (§5) · ὥσπερ αἰδουμένη. → ὥσπερ αἰδουμένη, ✱ · ἵνσ μή τις → ἵνα μή τις ✱ · τῇ παρ´ ἐαυτῆς καλλονῇ → παρ᾽ ἑαυτῆς ✱ · ἡ παγκόσμιος ἐπέσιη χἄρά → ἐπέστη χαρά ✱ · λήμην τὲ νεφῶδες → λήμην τὸ νεφῶδες ✱ · τὴν λαιιπηδόνα → λαμπηδόνα ✱ · καὶ Καὐτὸς → καὶ αὐτὸς ✱ (intrusive Κ) · κρύκτειν → κρύπτειν ✱ · ποικίλον τε καὶ διαλαμπῆ sound ✱ · εἰς κάλλος ἔένον → ξένον ✱ · Πᾶσί τε φυτοῖς sound ✱ · ἐπιεικῆ, ἢ πρὶν → ἐπιεικῆ, ἣ πρὶν ✱ · Νῆν θάλασσα → Νῦν θάλασσα ✱ · ἑπιζητεῖ → ἐπιζητεῖ ✱ · ἀήρ ἥπιον → ἀὴρ ἥπιον ✱ · ἑκτιναξάμενος → ἐκτιναξάμενος ✱ · ἐκκαλεῖται. φίλα → ἐκκαλεῖται, φίλα ✱ · τὸ πτερὸν ἄπονον ἀφαπλοῖ → ἐφαπλοῖ ✱ · διασημαίνοντα sound ✱ · ὅι´ ὡν νῦν μὲν → δι᾽ ὧν ✱ · ἀρχόμεθα χαὶ πρὸς → καὶ πρὸς ✱ · δι´ ἡν ἄλλος ἄλλῳ → δι᾽ ἣν ✱ · δείξῃ. καὶ τὰ πέρατα → δείξῃ, καὶ τὰ πέρατα ✱ (§5) · ἠς ἡ τρυφὴ → ἧς ἡ τρυφὴ ✱ · οὔτως ἀντιδῶμεν → οὕτως ἀντιδῶμεν ✱ · τῆς Κξένης → τῆς ξένης ✱ (intrusive Κ) · μονογενοῦς sound ✱ · δι´ ἦς ἐξ ἀνθρώπων → δι᾽ ἧς ✱ · πόῤῥω λιμώττοντας sound ✱ · ἀλλοτρίωσις και ζωῆς → καὶ ζωῆς ✱ · Ὅ μένων ἐν ἐμοὶ → Ὁ μένων ✱ · ποιείν οὐδέν → ποιεῖν οὐδέν ✱ · Ἐὰν μή τις μείν ἐν ἐμοὶ → μείνῃ ✱ · ἐβᾶήθη ἔξω → ἐβλήθη ἔξω ✱ · ὡς τὸ κλῆμα sound ✱ (third witness reads πλῆμα; Calfa is right) · συνάγονσιν → συνάγουσιν ✱ · τοῦ Υἱρῦ τοῦ ἀνθρώπου → τοῦ Υἱοῦ ✱ · τὸ αἴμα → τὸ αἷμα ✱ · ὧστερ γὰρ → ὥσπερ γὰρ ✱ · τῆς σαρκικης Ριίσταται ζωῆς → τῆς σαρκικῆς διίσταται ζωῆς ✱ · αἰωνίων, → αἰωνίων. ✱ (§5) · ΑΓʹ → ΛΓʹ ✱ (§4) · προήγαγε τόκον. → τόκον, ✱ · χαράς μὲν καὶ ἡμεῖς → χαρᾶς μὲν ✱ · Χαῆρε … Χαῇρε → Χαῖρε … Χαῖρε ✱ · τῇ μητρὶ τοῦ Θεοῶ → τοῦ Θεοῦ ✱ · Λεηθῶμεν → Δεηθῶμεν ✱ · καὶ ἦ ὑπερένδοξος → καὶ ἡ ὑπερένδοξος ✱ · καὶ ὁ ταύτης [0117] γάς· → καὶ ὁ ταύτης [0117] Υἱός· ✱ — the third witness prints καὶ ὁ ταύτης | Υἱός· καὶ εἴη at the same column turn and the Latin concurs (ejusque filius); Calfa's γάς is the mangled remnant of Υἱός. Restored silently under 13a clause 1; the column anchor is kept in its printed position, before the restored word · καὶ βεοῦ μητρὸς → καὶ Θεοῦ μητρὸς ✱ · σὺν τῷ ὰνάρχῳ → σὺν τῷ ἀνάρχῳ ✱.

8. The Latin seam bleed at chunk 0006 — confirmed, and nothing restored

The launch brief's warning is accurate and is confirmed against the file. src/pg-latin/ isidore-glabas-sermo-3/0006.md runs 8,668 characters including frontmatter; from character 6,434 it leaves Sermon III entirely (the brief's "roughly the last fifth, from ~6,625" measured the body without the header, and is the same seam). The break is visible in the Latin itself: after nunc et semper et in saecula saeculorum. Amen. it prints EDITORIS PATROLOGIAE MONITUM — J.-B. Migne's own notice questioning whether the next homily is Isidore's — and then SERMO IV. — In summe venerandam dormitionem immaculatissimae Dominae nostrae Dei Genitricis et semper virginis Mariae, followed by that sermon's §I and §II. The third witness's crop of the same leaves does the same thing, and after its own Ἀμήν. carries the next sermon's Greek head — printing it, in its own damaged way, as ΛΟΓΟΣ A', Εἰ τὴν πάνσεπτον κοίμησιν τῆς παναχράντου Δεσποίνης ἡμῶν Θεοτόκου καὶ ἀειπαρθένου Μαρίας (the A' is its misread of Δʹ; the Latin numbers it IV).

Nothing from that stretch is translated, restored, or treated as missing. Our Greek chunk 0006 ends exactly where Sermon III ends — at the doxology and Ἀμήν. inside column 0117 — and the Latin's continuation is evidence that the Greek is complete, not that it is truncated. This is the same page-atomic crop behaviour the pilot documented; the EDITORIS PATROLOGIAE MONITUM and the Sermon IV head are plate apparatus and text belonging to the next work, and they are left for whoever takes up Sermon IV.

9. Marker totals, ratio, and shape

[var:] 1 (0005 §27, Luke 1:49 Μεγαλεῖα) · [lat:] 0 (two candidates raised, both rejected with reasons — §3) · [sic:] 1 (0004 §22 ἐπαρνόρθωσιν, two-witness) · [ed:] 1 (0004 at [0105], unrecoverable words) · [d:] 0.

Column anchors 1:1 with the Greek in all three chunks — 0004: [0100] [0101] [0104] [0105] · 0005: [0108] [0109] [0112] · 0006: [0113] [0116] [0117]. No ## heads in the Greek and none in the English. verify-english-pg.mjs reports no error on 0004, 0005 or 0006 (the run fails only on 0003, which is Batch A's and not yet written).

Ratio: Greek 1043 / 1009 / 1152 words → English 1567 / 1506 / 1627 = 1.50 · 1.49 · 1.41, tracking the Glabas homily's 1.49 rather than the chronicle 1.25. 0006 runs lowest because a third of it is the catalogue of spring — short paratactic Νῦν … clauses where English is not longer than Greek.

Where the sermon ends. Chunk 0006's tail is the end of Sermon III. The doxology … νῦν καὶ ἀεὶ καὶ εἰς τοὺς αἰῶνας τῶν αἰώνων. Ἀμήν. falls inside column 0117, and both the third witness and the Latin show Sermon IV beginning below it on the same leaf. So this work ends mid-column 0117, which is why chunk 0006 carries a [0117] anchor and stops without a further one — the same shape as Sermon I's ending inside column 0040.

10. For the merge — items Batch A and the merger must settle

1. ὑπερφυής. Fixed here as supernatural for internal consistency; Sermon I is split between "supernatural" and "beyond nature" and is shipped. Flagged, not edited. This is the same shape as Sermon II's owed θαυμαστ- sweep and should probably be folded into it rather than raised separately.
2. The new ὑπερ- family (surpassingly pure / blameless / glorious) is derived from the governing table's own arithmetic rather than invented; if the merger prefers "above-" or "super-", the three must move together.
3. Σύλληψιν μονύεις at Luke 1:34 (Batch A's chunk 0003) — the third witness's crop of that leaf carries Σ0λ- … µτννεις, i.e. it too has a damaged verb there. That is a live two-witness question for Batch A and is named here only because I checked it while confirming the lemma was outside my range.
4. The [sic:] bar. This batch fires one [sic:] on two-witness concurrence without a leaf render, following Sermon II adjudication 2. If the merger wants that bar raised to require a leaf render, ἐπαρνόρθωσιν is the single site to revisit, and the fallback is a silent correction plus this crux entry.
5. συμποικιλτός vs ποικιλτής — an open letter-level disagreement between the two Greek witnesses (§5). Not urgent; a leaf render settles it.