Sermo II in Praesentationem B. Mariae Virginis

Per-batch logs: cruces-A.md (chunks 0000–0002) · cruces-B.md (0003–0004). 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 — it was built to govern the four-sermon SET. Extensions below are additions to it, never redefinitions. They govern Sermons III–IV too.

Batch A additions: παναμώμητος = all-unblemished · ἄμωμος = blameless · ἀκηλίδωτος = unspotted · παναγέστερος = all-holier · θεαυγέστατος = most God-radiant · θεοπτικώτατος = most God-beholding · θεοειδέστατος = most God-formed · θεολαμπής = God-shining · θεοτερπής = God-delighting · θεοπρεπής = God-befitting · θεόνυμφος = God-espoused · θεόλεκτος = God-chosen · θεοδόχος = God-receiving · διακωδωνέω = sound the bell · λαμπαδουχέω = bear torches · καινοτομέω = wrought newly · περιπέζιος = that which treads the ground.

Batch B additions: πάνσεμνος = all-august · μεγαλουργική = great-working · ἱεράρχης = hierarch · θεοειδεῖς ἐμφέρειαι = God-formed resemblances · θεουργικός/θεουργικά = God-working.

The sermon's two spatial terms are kept apart and MUST stay apart — ἄβατα = untrodden places · ἄδυτα = inner sanctuary. Verified across the merge: 9 Greek occurrences of ἄβατ- and 4 of ἄδυτ-, every one rendered on its own side of the line, in both batches independently. This is the sermon's whole subject; collapsing them would dissolve its argument.

⚑ RULED AT MERGE — παράδοξος / θαυμαστός are split, and Sermon I is the odd one out

παράδοξος = wondrous · θαυμαστός / θαυμάσιος = marvellous · θαῦμα = wonder · θαυμάζω = marvel. Batch B's assignment, adopted.

The evidence is not a preference: the two words stand side by side in §21 (τῇ παραδόξῳ … προόδῳ … τοῖς ἔτι παραδοξότερον ἐσομένοις … θαυμαστὰ μὲν ᾖ τὰ κατ᾽ αὐτήν). A panegyric that puts two near-synonyms in one period is building on their difference; one English word for both flattens the argument at the exact place it is being made. That is the genre rule (translation-style.md, High rhetorical homily) applied, not extended.

This does not contradict Sermon I, it completes it. Sermon I already rendered παράδοξος "wondrous" (παραδοξότατον = "most wondrous"), so that half is inherited; what Sermon I did was also give "wondrous" for θαυμαστός, in two places. Chunks 0000–0002 needed no unification — neither word occurs in them (checked; Batch B's merge flag is moot for that range).

⚠ OWED, NOT DONE — a tier-B sweep of Sermon I's two θαυμαστ- sites (src/english/isidore-glabas-sermo-1/0002.md, 0003.md). Sermon I is shipped and deployed, so its English is not edited here on a merge session's own authority. Flagged for Wilson. Until then the set is internally consistent within each sermon but not across the seam, and that is stated rather than hidden.

ADJUDICATIONS MADE AT MERGE

1. [ed:] at 0000 (section head Δʹ) — STANDS. Deliberately NOT converted to a data/calfa-patches/ entry, though Batch A offered that and the Joel precedent (patches 5/8/10/11/12) would seem to license it. Why not: a patch changes stored Greek and the attribution ladder therefore demands a plate check — letter-evidence, not sense-evidence. What we have is an OCR of the plate, and that OCR is itself damaged at this very line: the third witness reads A. ᾽Αλλὰ δὴ καὶ τὰ ποώην "Άγια ἁγίων, — ποώην for πρώην, A. for Δʹ. To patch, we would have to write letters no witness gives cleanly. The restoration is certain in substance and uncertain in letters, and [ed:] is precisely the instrument for that. A leaf render would license the patch; until one exists, the marker is the honest form. The substance is doubly confirmed: the third witness carries the words at the foot of col 0041, and Migne's Latin numbers the section IV (Quinimo priora illa Sancta sanctorum hanc…), whose hanc picks up the appellation that Calfa's next surviving words (νομίζω, διὰ τοῦτ᾽ ἔσχε τὴν πρόσρησιν ταύτην) govern — so the sentence construes across the join and nothing further is missing.
2. [sic:] at 0000 §3 θαυμαρτουγιῶν — STANDS, and it sets a precedent worth naming. Calfa AND the independent scan-Greek print the same ρτ/τρ metathesis. This is the first PG [sic:] fired on two-witness concurrence without a leaf render — which is exactly what ladder row 3 (both columns concur → the plate pair) was written to license. Recorded so the bar is not re-litigated per work.
3. Missing section Ηʹ (8) — logged, nothing supplied. The Latin's IX proves eight sections precede; our Greek yields seven marks. Section 8 falls inside the leaf gap in both columns (Calfa pp. 29–30 have no scan leaf), so there is no witness. Ladder row 4: no attribution, no guess, no [ed:] hole invented.
4. Unopened bracket at 0000 @0045 — carried asymmetric, correctly. The Greek source has 5 [ against 6 ]; the English has 7/8, the delta being exactly the two English-only markers. The asymmetry is transmitted 1:1 rather than silently balanced (Pattern 8), in a column that has no arbiter.
5. language: "grc" in the English frontmatter — KEPT. Every shipped PG work (sermo-1, nicetas, oecumenius, joel) does this; both agents kept it and both were right. The launch prompt's instruction to adjust it was mine and was wrong — recorded so the error is not re-issued to the next PG batch.
6. ἕως οὗ at B §23 rendered literally "so long as" though Allatius reads it causally (quando). Both Greek witnesses print it; the causal English would have read better and left no trace. Kept literal — the 7a case exactly.
7. The second semel per annum site (B, 0003) stays unmarked. One [lat:] records the divergence; a second marker on Allatius repeating his own supply would be marking the same fact twice.

⚑ The leaf gap is the real constraint on this work, and it is bigger than a numeral risk

Greek cols 0044 and 0045 are Calfa-only. Their Latin twins (0043, 0046) sit on the same absent leaves, so for that stretch there is no second witness of any kind — no twin, no third witness. Everything from §4 to §7 rests on our OCR alone. Nothing in that range is attributed to Migne's plate anywhere in this work's apparatus, and section 8's disappearance (item 3) falls inside it.

Marker totals after merge

[var:] 6 · [lat:] 1 · [sic:] 1 · [ed:] 1 · [d:] 0. Anchors 1:1 with the Greek in all five chunks; verify-english-pg clean.

The twin pass, per chunk — including where it found nothing

Batches A and B both collated polarity site by site against the Latin and reported no disagreement anywhere between the columns. Chunks 0000, 0001 and 0002 each yielded markers or logged divergences. Chunk 0004 was clean of [lat:], [var:] and polarity disagreement (ladder-row-1 OCR corrections only), and its full negation collation is itemized in cruces-B.md §3 so the null result is checkable rather than merely asserted. Chunk 0003 was not clean.

Scripture — the payload of a homily, per 7a″

Divergences from the received text, all carried as printed: Heb 10:19 prints ἐν τῷ ὀνόματι Ἰησοῦ for ἐν τῷ αἵματι (Latin concurs — the plate pair, not our file) · Ps 50:7 printed negated with an added μόνην, an authorial inversion carried in full · Luke 19:38 refitted feminine · Luke 2:30–32 passive ἡτοίμασται without σου · Isa 40:5 with an inserted address · Mk 10:14 printed with Matthew's ἡ βασιλεία τῶν οὐρανῶν and καὶ μὴ κωλύετε under Migne's own Mark citation (both columns concur — the plate pair's conflation). Seven further refits logged unmarked in cruces-B.md, incl. Luke 2:52 refitted to the Virgin and Heb 6:19–20 with its anchor rebound from hope to her grace. Counter-instance worth keeping: at Ps 17:7 the Greek is verbatim LXX and the Latin follows the Greek future, not the Vulgate — evidence that Allatius is not uniformly Vulgate-conforming, which bears on the no-marker rule earned in pilot §8 Q3.

Blind polarity read — 2026-08-01 (runbook step 4a; two Opus units, both columns)

Two independent blind units, neither having seen any cruces file, both declaring integrity intact. Unit 1: chunks 0000–0002. Unit 2: chunks 0002–0004. Chunk 0002 was deliberately in BOTH ranges — the overlap is the control on the whole pass, not redundancy.

Result: 4 ours-side defects, all in chunks 0000–0001, ALL FIXED. Chunks 0002–0004 clean on two independent reads.

⚑ Recall was measured — which is what makes the clean chunks trustworthy

Unit 2's three findings were all Calfa numeral damage that the translating agents had already caught and logged (ΠΗʹ→18, ΚΗ→22, ΚΜΗʹ→28). It rediscovered them blind, from the plate witnesses, with no access to the apparatus. Unit 1 likewise re-derived the [var:] sites already marked. So the "no defect" verdicts here are a measured null, not an unexamined one — the same control the 11065 read established. Chunk 0002 was read by both units and both found no ours-side defect.

The four defects — all invisible to any automated check

1. @0040 §1 — a printed negation dropped. ποῦ γὰρ ἄνθρωπος, ὅς, οὐδ᾽ εἰ μίαν βεβιωκὼς ἡμέραν εἴη … ῥύπου μένειν ἔξω πεπίστευται. We had "though he have lived but a single day" — οὐδέ rendered as a concessive, the "not even" gone. Latin: qui neque si unam tantum diem vixerit … sine macula esse creditur; third witness reads οὐδ᾽ too. The Job 14:4 LXX topos. Fixed → "not even if he has lived but a single day."
2. @0045 §7 — negation wrongly scoped, and the sentence contradicted itself. ἐπεὶ δὲ οὐχ ὁ βραχύ τι παρ᾽ ἀγγέλους ἥττων γεγένηται ἄνθρωπος, ἀλλ᾽ ὅ γε … τοὺς ἀΰλους ἐξαίρων. We had "man has become less than the angels not by a little" — i.e. much less — immediately followed by "but rather outtops the immaterial ones." The article ὁ makes βραχύ τι παρ᾽ ἀγγέλους ἥττων a quotation of Ps 8:6 / Heb 2:7, and οὐχ … ἀλλ᾽ ὅ γε is a not-X-but-Y frame: not the one a little lower than the angels, but the one who outtops them. The whole argument (this ἄνθρωπος is Mary; therefore her house must be loftier) rests on the elevation. Fixed. ⚠ This one sits INSIDE the 0044–0045 blind spot — no twin, no third witness. It rests on our Calfa OCR alone, which is however unambiguous here. Recorded as such; nothing about the plate is claimed.
3. @0048 §7 — predicate mis-bound, target reversed. οὐκ ἀσθενῆ τὴν Παρθένον πεποίηκεν ἔλεγχον, ἣ τοῦτον ἔτεκε is a double accusative: he made the Virgin who bore him a not-feeble refutation of the heretics. We had "he made no feeble refutation of the Virgin who bore him" — which reads as an objective genitive and points the refutation at her. The Latin is decisive and even supplies the nempe: invictum exhibebat argumentum illam nempe Virginem, quæ eum genuit. Fixed → "he made the Virgin who bore him no feeble refutation."
4. @0053 §13 — μόνην scoped wrong by English word order. οὐκ ἐν ἁμαρτίαις ἐκίσσησέ με μόνην ἡ μήτηρ μου. We had "my mother did not conceive me alone in sins," which in English reads "did not conceive only me" — implying others were not either, and dissolving the point, which is that Mary uniquely can say it (ἡ πάναγνος ἔχοι μόνη two lines above). Allatius saw the same trap and fronted it: Solam non concepit in peccatis me mater mea. Fixed** → "me alone did my mother not conceive in sins."

⚑ The durable lesson — genre predicted this too, and predicted it right

7a″'s rule is that genre predicts the failure mode. A chronicle's risk is the numeral (Joel, this same day: 35 polarity sites, zero ours-side defects, every historical defect a numeral). A high rhetorical homily's risk is the long periodic sentence — and all four defects here are one shape: a negation or a qualifier attached to the wrong element of a period. Not one is a mistranslated word. Three of the four produce fluent, plausible English, and #2 produces English that visibly contradicts itself one clause later and still survived drafting, self-review, and a full twin pass.

What this says about the twin pass: both translating agents ran it and both reported no polarity disagreement between the columns — and they were right, the columns do not disagree. The defects were between the Greek and our English, which the twin pass does not look at. A twin pass is not a polarity read and cannot replace one. Put this in every homily prompt: hunt the scope of the negation across the whole period, not the presence of the negative word.

### Recorded, not actionable (unit 1)

• @0057 §17: the comparative ὑψηλότερον τῶν φθασάντων binds to the perceiving in our English, to the illumination in Allatius. Both grammatically available, no polarity at stake — Allatius is a witness, not a standard. Left.
• Two blind-spot notes at §6 and §7 (τὸν ὑπὲρ νας ἄνθρωπον, possibly ὑπὲρ ἡμᾶς; and a damaged εἴγε ἐνῆν clause). Lexical/restoration, no witness available to settle either. Left unclassified, deliberately.
• §12 καλλίπαις construed "fair-childed"; Allatius speciosæ hujus puellæ. Both defensible, no polarity. Left.

The work-wide conventions section of src/english/isidore-glabas-sermo-1/cruces.md is binding here and is not restated. This file records (a) epithets and coinages new in Sermon II and the English fixed for each, (b) the scripture-divergence list, (c) the Latin-twin pass per chunk, (d) silent OCR corrections, (e) the one [sic:] and one [ed:] fired, and the [lat:] candidates rejected with reasons.

Witnesses. Greek = src/greek/isidore-glabas-sermo-2/ (Calfa). Latin twin = src/pg-latin/isidore-glabas-sermo-2/. Third witness = src/pg-greek-scan/isidore-glabas-sermo-2/ (the scan's own Greek column, independent of Calfa).

Frontmatter note. Each English chunk copies its Greek chunk's frontmatter verbatim, including language: "grc". The prompt said to adjust language, but all four English chunks of the Sermon I anchor keep grc, and the anchor is the binding format model; verify-english-pg.mjs passes. Flagged for the merge session — if the field is to be en, it is a one-line sweep across both sermons, not a Sermon II decision.

0. ⚠ THE LEAF GAP — Greek cols 0044 and 0045 are CALFA-ONLY

greekPlateAbsentPages = [29, 30]. On the four-column leaf layout of this work (Lat 0039 ‖ Gr 0040 · Gr 0041 ‖ Lat 0042 · Lat 0043 ‖ Gr 0044 · Gr 0045 ‖ Lat 0046 · Lat 0047 ‖ Gr 0048) the two absent leaves take out Greek 0044 and 0045 together with their Latin twins 0043 and 0046. Both crop files carry explicit [gap: …] markers at exactly those columns, which confirms the mapping.

Consequence, stated once and relied on throughout: everything in chunk 0000 from νομίζω, διὰ τοῦτ᾽ ἔσχε τὴν πρόσρησιν ταύτην (§4) to ἢ δήπουθεν ὑπὲρ τούτους (end of §7's long period) has no third witness and no Latin twin. Nothing in that stretch is attributed to Migne's plate; every reading there is logged as Calfa-only. Chunks 0001 and 0002 are fully covered by both verifiers.

One partial rescue: the Latin's col. 0047 opens with the tail of the sentence that began in the lost col. 0046 (appareret, et coelestem ejus originem et coelum propriam ejusdem esse sedem, utpote quae angelica conditione, imo supra angelicam conditionem creata sit), and Ballerini's note (54) on that column quotes the Greek itself — κατ᾽ ἀγγέλους δημιουργηθείσης and ἢ δήπουθεν ὑπὲρ τούτους. That plate-verifies the last clause of the gap range from Migne's own apparatus. It is the only such rescue.

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

Extending (never redefining) the Sermon I table. Every entry below is a word Sermon I did not use; the anchor's equivalences are carried unchanged (πάναγνος = all-pure · πανυπέραγνος = all-surpassingly pure · Πανάχραντος = All-immaculate · πανάμωμος = All-blameless · ἀμώμητος = unblemished · πανύμνητος = all-hymned · πολυύμνητος = much-hymned · Θεοτόκος = Theotokos · Δέσποινα = Lady · Δεσπότης = the Master · βασιλίς = Queen · παμβασιλίς = all-queen · θεόπαις = God-child · πολύτοκος = many-bearing · καλλίπαις = fair-childed · Κτίστης = Creator · Δημιουργός = Maker · πλάστης = Fashioner · δημιούργημα = created work).

⚑ The seven-member θεο- pile-up at 0001 §11 is this batch's epithet-flattening trap, and it is the exact analogue of the anchor's πολυύμνητος/πανύμνητος case. Migne's Greek runs θεαυγέστατον … θεοπτικώτατον … θεοειδέστατον … θεολαμπής … θεοτερπής … θεοπρεπής … θεόνυμφος καὶ θεόπαις. Allatius renders the middle three with splendor divinitus emicans · jubar Deo gratissimum · venustas Deum condecens — good Latin that keeps three of them apart, but he collapses θεοπτικώτατον ἔσοπτρον / θεοειδέστατον κάτοπτρον into one figure (expressissima Dei imago speculum Dei speciem ad vivum referens), losing the mirror/looking-glass doublet on which the sentence's rhythm turns. English keeps seven distinct compounds and both mirrors.

Register note carried from the anchor (capitalize a substantivized epithet where Calfa capitalizes, lower-case where it is adjectival): applied throughout. Hence "the All-pure one" at 0000 §8 (τὴν Πάναγνον) and 0001 §9/§11 (τῇ Πανάγνῳ, τὴν Πάναγνον), but "the all-pure one" at 0000 §1 etc. (ἡ πάναγνος). Likewise "the all-immaculate one" (0000 §2, lower-case in Calfa) against the anchor's capitalized "All-immaculate".

2. Scripture-divergence list — where Migne's printed Greek departs from the received text

This is the deliverable. Every place in 0000–0002 where the printed Greek differs from the received LXX/NT is named here. [var: …] fired only where the sense turns on it.

Marked [var:] in the English (5):

1. 0001 §9 — Heb 10:19–20. Migne prints ἣν ἐνεκαίνισεν εἰς τὴν εἴσοδον τῶν Ἁγίων ἐν τῷ ὀνόματι Ἰησοῦ; the NT has ἐν τῷ αἵματι Ἰησοῦ, "in the blood of Jesus." Two-witness concurrence — the Latin column reads initiavit in introitu Sanctorum in nomine Jesu (col. 0047 region, twin 0001 §IX frame), and Migne's own note cites Hebr. x, 19, 20. So the substitution is the plate pair's, not our file's (ladder row 3). Material: the letter to the Hebrews grounds the new way in the blood; Isidore's text grounds it in the name. Rendered as printed, marker names the NT reading.
2. 0001 §13 — Ps 50:7 (LXX), inverted and with μόνην added. Migne prints Οὐκ ἐν ἀνομίαις συνελήφθην, οὐκ ἐν ἁμαρτίαις ἐκίσσησέ με μόνην ἡ μήτηρ μου. LXX is affirmative and has no μόνην. This is a polarity inversion and it is the author's own — he announces it four words earlier (τὸ προφητικὸν ἐκεῖνο διαφυγεῖν, "to escape that prophetic saying"). The Latin concurs exactly (In iniquitatibus concepta non fui; et rursus, Solam non concepit in peccatis me mater mea), and Ballerini's note (61) builds an immaculist argument on it. Both negatives are carried into the English and neither is smoothed.
3. 0002 §18 — Luke 19:38, refitted to the feminine. Migne prints Εὐλογημένη ἡ ἐρχομένη βασιλὶς ἐν ὀνόματι Κυρίου· Εἰρήνη ἐν οὐρανῷ καὶ δόξα ἐν ὑψίστοις for the NT's masculine εὐλογημένος ὁ ἐρχόμενος ὁ βασιλεύς and the order ἐν οὐρανῷ εἰρήνη. Latin concurs (Benedicta quae venit regina … Pax in caelo). The refitting IS the homiletic move (the priest greets the Virgin with the Palm-Sunday acclamation); the English must not drift back to "Blessed is he that cometh."
4. 0002 §18 — Luke 2:30–32. Migne prints Νῦν εἶδον οἱ ὀφθαλμοί μου τὸ σωτήριον, ὃ ἡτοίμασται κατὰ πρόσωπον πάντων τῶν λαῶν … καὶ δόξαν λαοῦ Ἰσραήλ — passive where the NT has the second-person active ἡτοίμασας ("which thou hast prepared"), and without σου in both τὸ σωτήριόν σου and λαοῦ σου Ἰσραήλ. Latin concurs on the passive (quod paratum est) but restores the Vulgate's possessive in plebis tuae Israel — a Vulgate conformation in the Latin column, which by standing rule is convention, not divergence, and fires nothing. Marker names the NT reading.
5. 0002 §18 — Isa 40:5 (LXX), with an address inserted. Migne prints νῦν ὄψεται, κόρη, πᾶσα σὰρξ διὰ σοῦ τὸ σωτήριον τοῦ Θεοῦ. Migne prints the LXX himself in Ballerini's note (66): Isa. xl, 5, sec. LXX: Καὶ ὄψεται πᾶσα σὰρξ τὸ σωτήριον τοῦ Θεοῦ. Marked nonetheless, following the anchor's own precedent (its Ps 88:7 case was marked although Ballerini's note (9) adjudicated it on the page): the reader of our English does not see Migne's Latin apparatus.

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

• 0000 §7 — Ps 48:13/21 (LXX) παρασυνεβλήθη τοῖς κτήνεσι τοῖς ἀνοήτοις, καὶ ὡμοιώθη αὐτοῖς is verbally the LXX, woven into the homilist's own clause. No divergence.
• 0002 §16 — Luke 2:22. ἀνήγαγον, φησίν, εἰς Ἱεροσόλυμα παραστῆναι Κυρίῳ for the received transitive παραστῆσαι τῷ κυρίῳ ("to present him to the Lord"). The Latin has the transitive (ut sisterent eum Domino) — i.e. the Latin agrees with the received text against Migne's Greek. This is the closest thing in the batch to a [lat:], and it is not fired: the difference is one of voice, the object is supplied by the same sentence, and no reader of either column takes away a different fact. Rendered intransitively, "to stand before the Lord." Also note τὸν ἐμὸν Ἰησοῦν for αὐτόν — the homilist's own weave, which the Latin keeps (Jesum meum).
• 0002 §18 — Ps 44:11–12 (LXX). ἐπιλάθου τοῦ οἴκου τοῦ πατρός σου drops τοῦ λαοῦ σου καί; and ὁ βασιλεὺς τοῦ σοῦ κάλλους ἐπιθυμήσει reorders ἐπιθυμήσει ὁ βασιλεὺς τοῦ κάλλους σου. A truncation plus word order, not a variant.
• 0002 §19 — Ps 42:4 + Ps 21:10–11 (LXX), welded. Εἰσελεύσομαι … πρὸς τὸν Θεὸν τὸν εὐφραίνοντα τὴν νεότητά μου is Ps 42:4 verbatim; ὅτι σὺ ἡ ἐλπίς μου ἀπὸ μαστῶν τῆς μητρός μου· ἐπὶ σὲ ἐπεῤῥίφην ἐκ μήτρας, ἀπὸ γαστρὸς μητρός μου, Θεός μου εἶ σύ welds Ps 21:10b–11 and reads ἀπὸ γαστρός for the LXX's ἐκ κοιλίας. Migne cites both psalms on the plate, flagging the weld himself. Rendered as the weld stands.
• 0002 §20 — Luke 2:40. ηὔξανε … καὶ ἐκραταιοῦτο πνεύματι, πληρούμενον σοφίας, καὶ χάρις Θεοῦ ἦν ἐπ᾽ αὐτῷ. πνεύματι is the Byzantine/TR reading, not a divergence. σοφίας for the dative σοφίᾳ, and ἐπ᾽ αὐτῷ for ἐπ᾽ αὐτό, are case-level. Worth naming: the Latin feminizes it — et gratia Dei erat in illa — where the Greek keeps the neuter agreeing with παιδίον. Not fired: the referent is the same person either way. English keeps the neuter, "upon it."
• 0002 §16 — Ps 88:16 (LXX) Μακάριος ὁ λαὸς ὁ γινώσκων ἀλαλαγμόν and §18 Ps 22:6 κατοικεῖν … ἐν οἴκῳ Κυρίου εἰς μακρότητα ἡμερῶν: verbally as printed in LXX, woven.
• 0002 §18 — Zeph 3:14 (LXX) Χαῖρε σφόδρα, θύγατερ Σιών: exactly the LXX, and Migne adjudicates it on the page — note (65): Sic versio LXX edita ab Aldo: Χαῖρε σφόδρα, θύγατερ Σιών. Vulgatus noster: Lauda filia Sion. No marker. Rendering decision, carried from the anchor: χαῖρε is englished Rejoice, not Hail, because the next clause is δι᾽ ἧς ὁ Κτίστης … χαίρει … χαρήσεται … τὴν χαράν σοι … μηνύσει — a four-fold χαρ- figure that "Hail" would destroy outright.
• 0001 §13 — Luke 1:49. ὧν μοι μεγαλεῖα ἐποίησεν ὁ δυνατός for the received ἐποίησέν μοι μεγάλα. Latin magnalium … magna. Word-choice, sense unchanged.

3. The Latin-twin pass, chunk by chunk — including whether each was clean

Method as specified: draft first, then one deliberate pass against the twin hunting (1) OCR damage in our Greek, (2) different asserted facts, (3) polarity disagreement.

• Chunk 0000 — NOT clean. One [sic:] (§4 below), one [ed:] (§5), ~25 silent OCR corrections (§6). No [lat:]. Polarity collation, every site: οὐδέποτ᾽ ἄν / nunquam · οὐδ᾽ εἰ μίαν … ἡμέραν / neque si unam tantum diem · ῥύπου μένειν ἔξω / sine macula esse · οὐδὲν … διοίσει / nihil … distare · οὐδ᾽ ἂν ἓν … οὐδέν, ὃ μὴ … ἐντίθησιν / nihil omnino sit, quod … non ingerat (the double negative carried in both columns and in the English) · οὐδὲν ἀληθῶς ἕτερον / nil aliud · χρῆσιν οὐδενὶ παρέχων / nullum cuipiam commodum · μὴ ὅτι τοῦτ᾽ ἐγεγόνει / illum non reipsa factum fuisse hominem · οὐκ ἐνῆν / haud par fuisset · ἥκιστα ἔπτη / haudquaquam evolavit. No disagreement anywhere. Numerals: none but section marks (§7). Free-rendering, named so a later pass does not re-raise: οὐκ ἀσθενῆ … ἔλεγχον → invictum … argumentum (litotes flattened to a positive — Allatius's norm); τῷ γαληνῷ τοῦ Σωτῆρος → affabili Salvatoris benignitate; περιπεζίου παντὸς ἀνωτέρῳ → terrenis omnibus sublimius.
• Chunk 0001 — NOT clean. Two [var:]. No [lat:]. Polarity: οὐκ ἦν … τυχεῖν / non expediebat · οὔτ᾽ ἦν ἱκανὸς … καὶ εἶχεν ἄλλως / in sacris penetralibus morari non poterat, et … interpretari debuit — the broken correlative (οὔτε … καί for οὔτε … οὔτε) stands in BOTH columns and is carried into the English ("who neither was fit … and was otherwise able"), exactly as the anchor's Ῥουβίμ case was · οὐχ … μόνον, ἀλλὰ καί / non modo … sed · οὐ τῆς ἴσης … θεοπτίας / haud pari mensura · οὐδ᾽ ὅσον εἰπεῖν ἔνεστιν / quantum neque verbis eloqui fas est · μήτ᾽ ἐκείνους … μήτε τὴν μακαρίαν / non aliam … non alios · Οὐκ ἐν ἀνομίαις … οὐκ ἐν ἁμαρτίαις / non fui … non concepit · τῶν οὐκ εἰκότων / incongruum · οὐ μετεῖχε τραπέζης / haud humana alebatur mensa · Οὐδὲ γὰρ ἦν οἴκοι μένειν τῶν ἀναγκαίων / ut domi maneret nequaquam necesse fuit. No disagreement.* Numerals: τριετία / triennium and τριετής / triennis* — agree, no plate check needed.
• Chunk 0002 — NOT clean. Three [var:]. No [lat:]. Polarity: οὐκ ἦν ἂν τῶν ἀκολούθων … ὑπερβῆναι / haud consentaneum fuisset … praetergrederentur · μὴ διαπεσεῖν εὖ πεποιήκασι / recte de promissione implenda curaverunt (recast, no polarity flip) · οὐδὲν ἦν, ὃ τοῦ ἱεροῦ μὴ τῇ Παρθένῳ ἀνέῳξε / nulla templi pars fuit, quam Virgini perviam non faceret — double negative carried in both and in the English · ἀγγέλους ἀγνοεῖν / angelis incompertum · οὐδ᾽ εἰς φρονοῦσαν ἡλικίαν τελοῦντες ὑπεριδεῖν ἔχουσι / e qua aegre avelli se sinunt vel qui jam matura sint aetate · οὐκ ἂν ἱκανὸς εἰπεῖν εἴη / vix effari quispiam poterit. No disagreement.* Numerals, all agreeing: τριετίας / ad triennium · τὸ τοῦ ἔτους δίκαιον μέτρον … ἡ τριετία / justam temporis mensuram … triennium · τρία … ἔτη / trimula · τοῦ ἐνιαυτοῦ ἅπαξ / semel tantum in anno*. No numeral [lat:].

Chunks on which the twin pass found nothing at all: NONE. All three yielded at least one marker or divergence, so no chunk in this batch is reportable as clean. The absence of a clean chunk is itself the honest result of the pass.

Pronoun-itacism check (ἡμεῖς/ὑμεῖς), the class the paired pilot flagged: three sites — ῥύπος εἰδότες ἡμεῖς (0000, Lat. quarum conscii nobis sumus), ὁδὸν ἡμῖν ὑποδείξῃ (0001, Lat. viam nobis ostenderet), τῶν πρὸ ἡμῶν (0001, Lat. e majoribus). All agree across the columns. No [lat:] in this class.

4. [sic:] — one fired, and why the third witness makes it safe

0000 §3 — θαυμαρτουγιῶν for θαυματουργιῶν. Calfa reads καὶ τῶν πάλαι θαυμαρτουγιῶν παραδόξων ἔκθεσίς τε καὶ διασάφησις ἐτελεῖτο. The scan's own Greek column, independent of Calfa, reads the same letters — θαυμαρτουγιῶν — in the same place. Two independently-transmitted OCRs of the same plate agreeing on a ρτ/τρ metathesis is not an OCR-confusion profile; it is what the plate prints. The Latin gives the intended sense (admiranda illa prodigia, quae olim perpetrata fuerant), so the word meant is θαυματουργιῶν, "wonder-workings."

Carried into the English untranslated, in italics, inside [sic: …] (Patterns 7, 10, 12): "the setting forth and elucidation of the marvellous [sic: θαυμαρτουγιῶν] of old." The conjecture lives here, not in the text.

⚑ For the merge session: this is the first [sic:] of the PG paired pilots, and it is fired purely on scan-Greek/Calfa concurrence with no leaf render. If the merge session judges that bar too low for a public claim about Migne's plate, the fallback is a silent correction plus this crux — but note that the standing rule forbids attributing a defect to the plate from our files alone, and two independent witnesses of the plate are precisely what §4a says is no longer "our files alone." Recorded so the decision is made deliberately.

5. [ed:] — one fired: the lost section head Δʹ

0000, at the foot of Greek col. 0041. Calfa runs … ἡ πάναγνος, φημ, καὶ φωτοειδεστάτη Θεοῦ νύμφη. [0044] νομίζω, διὰ τοῦτ᾽ ἔσχε τὴν πρόσρησιν ταύτην — a column break at which the sentence resumes with a bare νομίζω that has no subject and no section number, in a text whose sections have run Αʹ Βʹ Γʹ.

The scan settles it. Its Greek column ends col. 0041 with

Δ΄. Ἀλλὰ δὴ καὶ τὰ πρώην Ἅγια ἁγίων,

and the Latin twin's col. 0042 ends with IV. Quinimo priora illa Sancta sanctorum hanc — the section number in both witnesses. Restored, the sentence construes perfectly: Ἀλλὰ δὴ καὶ τὰ πρώην Ἅγια ἁγίων, νομίζω, διὰ τοῦτ᾽ ἔσχε τὴν πρόσρησιν ταύτην…

This is Pattern 13a clause 2 — a reader-visible structural unit (a division head) lost from our source and recoverable from the plate — so it is restored with an [ed:], not silently. Without the marker the sermon would read as running from §3 to §5, and nobody can miss what they cannot see was ever there.

Recommended follow-up (not done here, since it changes stored Greek): a data/calfa-patches/ entry restoring Δʹ. Ἀλλὰ δὴ καὶ τὰ πρώην Ἅγια ἁγίων, at the foot of col. 0041. The ladder requires a plate check for a patch; the plate check is in hand and cited above (two witnesses).

6. Silent OCR corrections — classes first, then per site

Class 1 — the intrusive Κ, already attested seven times across three works in this Calfa volume. Six more here, all corrected silently: Κεἶκε πέδαις → εἶκε πέδαις ✱ (0000 @0048) · ΚΖʹ → Ζʹ (0000 @0045, §7 — see §7 below) · καταβεβηΚκότες → καταβεβηκότες (0001) · αἰτίαν Κὁμιλίαν → αἰτίαν ὁμιλίαν (0001) · τὸν Κἐμὸν Ἰησοῦν → τὸν ἐμὸν Ἰησοῦν ✱ (0002) · θεραπείας Κλλης → θεραπείας ἄλλης ✱ (0002). A near relative in the same class: Χηρημάτων → ληρημάτων ✱ (0000 @0048). The class now runs to fourteen attested sites in this volume — worth a corpus note, not worth per-site marking. (✱ = plate-verified against the scan's Greek column.)

Class 2 — split and run-together type, our source's not the plate's, therefore corrected rather than carried: τὸ ὕ, ος → τὸ ὕψος ✱ (0000) · δὲνκαθάπερ → δὲ καθάπερ (0000) · ἐκεῖ. α → ἐκεῖνα (0000) · μεμόνην → με μόνην (0001) · προσιστα μένου → προσισταμένου ✱ (0002) · ἠλλά- ἔατο → ἠλλάξατο ✱ (0002) · ηΧξαλεκαὶ → ηὔξανε καὶ ✱ (0002).

Words lost mid-sentence, restored silently from the plate (Pattern 13a clause 1): 0000 §3, τοῦ θείου μᾶλλον ‹ἐκείνου τεμένους ἁγιαστικὴν ἕξιν ἢ κλῆσιν αὐχοῦν› — seven words dropped by Calfa at a line break, present in the scan's Greek column and confirmed by the Latin (nil aliud reperire erat quod … sanctificandi vi, aut sancti appellatione gloriari posset). Without them the clause has no predicate at all. 0001 §12 tail, the terminal full stop after τῷ δεδωκότι (the scan prints it; Calfa runs straight into ΙΓʹ), restored.

Per site, chunk 0000 (✱ = plate-verified): τῆ παναχράντου → τῆς παναχράντου ✱ · ἐχθύων → ἰχθύων ✱ · παναμώμω → παναμώμῳ ✱ · Πν´ ὅτι → ἵν᾽ ὅτι · οὐδεν, → οὐδέν, ✱ · τρόπαιᾳ → τρόπαια ✱ · ὁ χωρος → ὁ χῶρος ✱ · ἐκρύβη τα → ἐκρύβη τε ✱ · Ἅγκα ἁίων → Ἅγια ἁγίων ✱ · φημ, → φημὶ, ✱ · ὰξίας → ἀξίας · οἰκουοήσοντα → οἰκήσοντα · πρὶν τοῦτα → πρὶν ταῦτα · τεχνητῳ → τεχνητῷ · τοόπον → τρόπον · προτέοων → προτέρων · οἷκητήριον → οἰκητήριον · τἰσεῖν → εἰπεῖν · παρασυνεβᾶήθη → παρασυνεβλήθη · κτήοεσι → κτήνεσι · ἀκάγκῃ → ἀνάγκῃ · νεών → νεὼν · τοὺς ἀθλους → τοὺς ἀΰλους (the same ἀ-ΰ class as Sermon I's ἀθλου δημιουργίας → ἀΰλου) · προδεδέόμηται → προδεδόμηται · ἐνἀοτάτοις → ἐνδοτάτοις · εἶκεν ἄν → εἶχεν ἄν · ὁ σώςων → ὁ σῴζων · χοῖκῶν → χοϊκῶν · ἀῖγέλους → ἀγγέλους ✱ (plate-verified from Ballerini's note (54), which quotes the Greek) · σωτήριός εε → σωτήριός τε ✱ · γητνων → γηΐνων ✱ · τοῦ χυοὸς → τοῦ χοὸς ✱ · ὡς εἰπεῖ → ὡς εἰπεῖν ✱.

Per site, chunk 0001 (all ✱ unless noted, the scan covering this chunk entire): ἡς → ἧς · Ἀγα → Ἅγια · ὁδὺν → ὁδὸν · ἑνεκαίνισεν → ἐνεκαίνισεν · σεμνῳ → σεμνῷ · τ´ ἐσφραγισμένον → τὸ ἐσφραγισμένον · Ἃγμα ἀγίων → Ἅγια ἁγίων · οἱκείοις → οἰκείοις · ἰσραρχῶν → ἱεραρχῶν · οὐδ᾽ ἂσον → οὐδ᾽ ὅσον · ἐπίκεινα → ἐπέκεινα · θεαυγέστατον sound · ἡγήοασθαι → ἡγήσασθαι · θαυμάσισς → θαυμάσιος · δικαιοτάτον → δικαιότατον · οἴαν → οἵαν · κἀκεῖνεν → κἀκεῖθεν · λγάπησε → ἠγάπησε · τούτον → τούτων · Ῥδει γὰρ → Ἔδει γὰρ · ἄκοον → ἄκρον · ἧς τὲ σεμνὸν → ἧς τὸ σεμνὸν · ταῖδα παοὰ → παῖδα παρὰ · συμπεριειληκμένου → συμπεριειλημμένου · μεγαλείον ἐτοίησεν → μεγαλεῖα ἐποίησεν · εὖ ἔον → εὖ ἔχον · συνέὃραμεν → συνέδραμεν · Ἰνʹ → ΙΔʹ (§7) · εόπαις → θεόπαις · περαναι → περᾶναι · ὕαύαστα. → θαύματα, · λουτοὸν. ὧ → λουτρόν, ᾧ · τιτθης → τιτθῆς · τὰ δυτα → τὰ ἄδυτα · ἔ,η → ἔφη · δλης φρικτῇς → ὅλης φρικτῆς · οἰκητήοιον → οἰκητήριον · οἴὔκοι → οἴκοι · κακοπαθεῖοθαι → κακοπαθεῖσθαι.

Per site, chunk 0002: δʹ ἂν → δ᾽ ἂν · διὰ σαῦτα → διὰ ταῦτα ✱ · πεπουήκασι → πεποιήκασι ✱ · βασιλίβος → βασιλίδος ✱ · ἱἔμελλε → ἔμελλε · ἐπάσαντες → ἐπᾴσαντες · πα αστῆναι → παραστῆναι ✱ · ᾗεσαν → ᾔεσαν · προειεῆσθαι → προειρῆσθαι ✱ · πυραὸν → πυρσὸν ✱ · ὁμνούσας → ὑμνούσας · οίονεί → οἱονεί · σκευος → σκεῦος ✱ · ἐαυτοῦ → ἑαυτοῦ · καρὰν → χαρὰν ✱ · ΠΗʹ → ΙΗʹ (§7) · Εὐλόμένη → Εὐλογημένη ✱ · Ελρήνη → Εἰρήνη ✱ · ἣνὁ βλέπων → ἣν ὁ βλέπων · θόγατερ → θύγατερ ✱ · θψούμενον → ὑψούμενον ✱ · τοῦθχ → τοῦθ᾽ ✱ · ἣτοίμασται → ἡτοίμασται ✱ · ἀποκάλιψιν → ἀποκάλυψιν ✱ · Ἡσατου → Ἡσαΐου ✱ · ἐπιθυμνσει → ἐπιθυμήσει ✱ · ἡρετίσατο → ᾑρετίσατο ✱ · νομίνεται → νομίζεται ✱ · Ἵι τοίνυν → Ἴθι τοίνυν ✱ · λαμπαλουχοῦσαι → λαμπαδουχοῦσαι ✱ · ψήφῳ συνάδειν → ψήφῳ συνᾴδειν · εἰπεν ἀν → εἶπεν ἄν ✱ · Δαβδ → Δαβίδ ✱ · εἰφραίνοντα → εὐφραίνοντα ✱ · Θεός μου εἰ σύ → Θεός μου εἶ σύ ✱ · ἰερου → ἱεροῦ ✱ · ἐστίας → ἑστίας · ἴτη → ἔτη ✱ · παιλίον → παιδίον ✱ · πιεύματι → πνεύματι ✱ · πληΕούμενον σοφίπς → πληρούμενον σοφίας ✱ · Πν ἐπ´ αὐτῷ → ἦν ἐπ᾽ αὐτῷ ✱. Also εἴρηται,, (double comma, in both Calfa and the scan) normalized to one — OCR noise on both sides, not a plate mark.

7. Section numerals — the sequence, and ONE UNRESOLVED

Rendered as arabic + period, per the anchor. Sermon II's sections run continuously from Sermon I's, restarting at Αʹ.

⚠ UNRESOLVED, and left deliberately unmarked — the missing Ηʹ (8). The Latin's IX at col. 0047 fixes that eight sections precede chunk 0001. Our Greek yields only seven marks (1–7). Section 8 is absent from every witness we hold: Calfa prints no Ηʹ; the scan has no leaf for Greek col. 0045, where it would fall; the Latin has no leaf for its twin col. 0046, where VIII would fall — and the Latin's col. 0047, which we do have, runs from the tail of §7 straight to IX with no VIII in it. So this is a ladder row 4 case: both witnesses fail together at the same spot, and the arbitrating leaf is the one the scan lacks. Per the leaf-gap rule the candidate is logged here and nothing is marked in the English: no [ed:], because we cannot demonstrate a loss as against Migne's own numbering skipping, and no restoration, because there is nothing to restore from. The likeliest position, on sense, is at Ὥσπερ γὰρ οὐ κατὰ τοὺς λοιποὺς τῶν ἀνθρώπων… (0000 @0045). If the leaf for Calfa p. 30 is ever obtained, this is the first thing to check — and it is a cheaper check than it looks, because the same leaf would also settle §7's ΚΖʹ.

8. [lat:] — NONE FIRED, and the candidates weighed

Zero [lat:] markers in this batch. Stated positively because it bears on pilot gate question 3 (how many divergences per work, and does Allatius's freedom generate spam). Across ~3,000 Greek words the Latin column and the Greek column assert no different fact anywhere — no numeral, name, negation or polarity divergence, and no added or missing clause that changes what a reader takes away. Candidates weighed and rejected:

• 0002 §16, Luke 2:22 παραστῆναι Κυρίῳ vs ut sisterent eum Domino. The strongest candidate: the Latin has the transitive of the received text where the Greek prints the intransitive. Rejected as a difference of voice, not of fact — §2 above.
• 0001 §10, ὁ ἅπαξ … εἰσιὼν ἱεράρχης vs qui semel per annum … pedem inferebat. The Latin adds "per annum." Rejected: the Greek itself supplies τοῦ ἐνιαυτοῦ ἅπαξ two chunks later (0002 §18), so Allatius is filling from the same text, not asserting otherwise.
• 0002, τὸ τοῦ ἔτους δίκαιον μέτρον vs justam temporis mensuram. Rejected: Allatius smooths an odd genitive; both columns then gloss the measure as triennium / ἡ τριετία, so the fact is identical.
• 0002 §20, Luke 2:40 ἐπ᾽ αὐτῷ vs in illa. The Latin feminizes the neuter. Rejected: same referent, and it is the Latin conforming its pronoun to the application the Greek makes in the same breath.
• 0000 §8, οὐκ ἀσθενῆ … ἔλεγχον vs invictum … argumentum. Litotes flattened to a positive. Textbook Allatius freedom; the English keeps the litotes ("no feeble refutation").

Free renderings named so a later pass does not re-raise them: speciosae hujus puellae for καλλίπαιδος (the same flattening the anchor caught, and the reason "fair-childed" is fixed) · expressissima Dei imago speculum… collapsing two of the seven θεο- compounds (§1) · ineffabili modo for καινότατα · librum legendi gnaro for εἰς πεῖραν ἥκοντι γραμμάτων · evidentissime proclamabat for διακωδωνοῦντος · aegre avelli se sinunt for ὑπεριδεῖν ἔχουσι. Scripture in the Latin column conformed to the Vulgate is convention, not divergence, and fires nothing: plebis tuae Israel (Luke 2:32) is the worked instance here — see §2 no. 4.

9. Construal cruces — printed reading kept, conjecture here

• 0000 @0045 — the unopened bracket. Calfa prints … τὴν δίαιταν ἔσχεν [ἄ]γγελος γὰρ ἦν οὐρανόθεν λειτουργῶν τῇ Παρθένῳ τὴν ἀπόῤῥητον ἐκείνην τροφήν] — a closing bracket with no opening bracket. Migne plainly sets the clause as a parenthesis and our source has lost the [. Not supplied (Pattern 8: never supply a mark, and the anchor's unclosed-guillemet precedent at Sermon I 0001 rules the mirror case the same way); the English carries the closing bracket alone, at the same point. This column is in the leaf gap, so neither the scan nor the Latin can arbitrate, and a leaf render would settle it. Flagged for the merge session as the one place in the batch where a reader may take a source defect for ours.
• 0000 @0045 — τὸν ὑπὲρ νας ἄνθρωπον. νας is not a word. Read νοῦν ("the human being beyond understanding"), which the parallel τὸ καινὸν δημιούργημα τοῦτο in the same clause supports. Calfa-only column: no Latin, no scan. The letter is unresolved; the sense is not in doubt.
• 0000 @0045 — οὕτω καὶ παρὰ τοὺς λοιποὺς ἀγγελικὴν εἶχεν ἄν τις ἑστίαν … εἴγε ἐνῆν, ὁ σῴζων ἠνάγκαζε λόγος ἀπολαβεῖν. The syntax does not resolve cleanly: εἶχεν ἄν τις … ἀπολαβεῖν wants an infinitive complement, and ὁ σῴζων … λόγος ἠνάγκαζε sits in asyndeton. Rendered as closely as the printed words allow, with an em-dash for the break, rather than smoothed into a single period. Calfa-only column; not repaired.
• 0000 @0048 — οὗπερ ἥκιστα πέπτη. πέπτη is not a word — and Migne says so himself. His note (98) on the facing column reads Ita codex, ut videtur, pro ἔπτη ("so the codex, it seems, for ἔπτη"). The Latin renders accordingly (E quo quidem limo haudquaquam evolavit). The plate has already adjudicated it, so no [sic:] of ours is stacked on top (Pattern 12's "where Migne prints his own note, do not add ours"); the English reads "from which she by no means flew away," and the reading is recorded here.
• 0000 @0045 — μόνον δικαίως ὡς ἑαυτοῦ δεσπότην, masculine, of the Virgin (τὴν θεόνυμφον … βασιλίδα). Pattern 9: Greek grammatical gender has no English exponent here, and the anchor already fixes Δεσπότης = "the Master." Rendered "as its own master." No emendation to make or undo; logged so a later sweep does not churn it.
• 0002 §16 — τὸ μὲν τὸν πατέρα, τὸ δὲ διὰ τὴν μητέρα. The first limb lacks the preposition the second and third have (διά). Both Calfa and the scan print it so — two-witness concurrence, ladder row 3, therefore the plate's ellipsis, not our file's loss. The Latin repairs it (illi quidem intuitu patris). Rendered with the ellipsis kept — "these for the father, these on account of the mother" — under Pattern 7. Not a [lat:]: Allatius is supplying syntax, not a different fact.
• 0002 §16 — φθάνει τὰ Ἱεροσόλυμα … προπλέξαντες. Singular verb with neuter plural subject, then masculine plural participles, and the same anacoluthon recurs at §16's ὁ νοητὸς … διάκοσμος … δεικνύντες. Pattern 9: no English exponent, no emendation. Rendered with the sense the participles carry; logged.
• 0001 §13 — ἀνθέων πᾶσαν εὐωδεστάτων κάτω τεθεὶς πέφηνε χάριν. πέφηνε with a bare accusative χάριν will not construe as a comparison without supplying a preposition. Rendered "has appeared beyond all the grace of the most fragrant flowers," which is the least that the printed words allow; the Latin has the participial exsuperans, i.e. it repairs the construction. The choice is recorded rather than made silently.
• 0001 §12 / §13 — οὔτε … καί, twice. The broken correlative at οὔτ᾽ ἦν ἱκανὸς … καὶ εἶχεν (§9) and οἵαν οὔτε Θεὸς παρορᾷ, καὶ ὡς ὀσμὴν … ὁρᾷ (§12) is the same class the anchor logged at Sermon I 0002 (Ῥουβίμ). Kept broken in the English ("neither was fit … and was otherwise able"); the Latin repairs both.
• 0000 §1 — ποῦ γὰρ ἄνθρωπος, ὅς, οὐδ᾽ εἰ μίαν βεβιωκὼς ἡμέραν εἴη … προσομιλεῖν ἄν τοτε … φανείη; ἄν τοτε (so in both Calfa and the scan) read as ἄν ποτε. The underlying allusion is Job 14:4 sec. LXX, which Migne cites on the facing column; it is an allusion, not a quotation, so no [var:].

10. Ratio and shape

Greek 1024 / 1011 / 1022 words → English 1610 / 1555 / 1504, i.e. 1.57× · 1.54× · 1.47×. Tracking the anchor's 1.49× homily ratio (chronicle 1.25 · lemma-gloss 1.35 · dialogue 1.37 · homily 1.49); the slightly high 0000 is the epithet pile-ups and the [ed:]/[sic:] prose, neither of which compresses.

verify-english-pg.mjs isidore-glabas-sermo-2 → 5 chunks, anchors + frontmatter + ratios + dedupe clean. Column anchors reproduce 1:1 in order: 0000 five ([0040] [0041] [0044] [0045] [0048]), 0001 four ([0049] [0052] [0053] [0056]), 0002 three ([0057] [0060] [0061]).

Marker totals for the batch: 1 [sic:] · 1 [ed:] · 5 [var:] · 0 [lat:] · 0 [d:] · 0 [n:].

11. For the merge session to adjudicate

1. language: "grc" in the English frontmatter — anchor-conformant, prompt-divergent. One-line sweep either way, across both sermons together.
2. The [sic:] at 0000 §3 (θαυμαρτουγιῶν) — first PG [sic:] fired on scan/Calfa concurrence without a leaf render. §4 states the case both ways.
3. The missing section Ηʹ — unmarked by design, ladder row 4. §7.
4. The unopened bracket at 0000 @0045 — carried asymmetric under Pattern 8, in a column with no third witness. §9.
5. A data/calfa-patches/ entry restoring Δʹ. Ἀλλὰ δὴ καὶ τὰ πρώην Ἅγια ἁγίων, at col. 0041, which would let the [ed:] at §5 be withdrawn in the same commit (Pattern 13's own rule: the note exists only as long as the hole does). Plate check is in hand.
6. Epithet-table merge — §1 extends, never redefines. The παναμώμητος / πανάμωμος / ἀμώμητος / ἄμωμος quartet is the entry most at risk of being collapsed by a later agent working from the Latin, which has only immaculata / sine macula.

Chunks 0003 (cols 0064–0068) and 0004 (cols 0069–0072, the sermon's last chunk, ending at the doxology and Ἀμήν). Chunks 0000–0002 were translated concurrently by a peer agent; nothing here touches their files.

The work-wide conventions section of src/english/isidore-glabas-sermo-1/cruces.md is binding and is not restated. Its Marian epithet table, maker-word table and coinage rule govern this sermon; §1 below records only what Sermon II adds.

Witnesses. Greek source = src/greek/isidore-glabas-sermo-2/ (Calfa). Latin twin = src/pg-latin/isidore-glabas-sermo-2/0003–0004.md (cols 0063–0067, 0070–0071). Third witness = src/pg-greek-scan/isidore-glabas-sermo-2/ — the scan's own Greek column, independent of Calfa; used for every plate check below, so no reading here rests on our files alone.

Frontmatter. Copied verbatim from the Greek chunks, language: "grc" included — matching the shipped PG English convention (sermo-1 anchor, nicetas, oecumenius all keep grc in the English chunk). Flagged in case the merge session wants the field flipped corpus-wide; it should not be flipped for this work alone.

1. Epithets and coinages — additions to the binding table

New in this batch; one English each, never interchanged:

Carried unchanged from the binding table and used here: πάναγνος / ἡ Πάναγνος = all-pure / the All-pure one · πανύμνητος = all-hymned (0004, the sermon's last epithet) · Παρθένος = Virgin · Δέσποινα, δέσποινα = Lady · Δεσπότης = the Master · βασιλίς = Queen · ἡ μακαρία = the blessed one · Πλάστης = Fashioner · δημιουργός = Maker · ποίημα = work · ἱστουργία = weaving (0004 §27, φυτῶν … ἱστουργία = "the weaving of plants" — the anchor's own coinage, and the Latin again drops the metaphor, ornatus).

⚑ Two Latin flattenings that the epithet rule caught, both worth naming so a later pass does not read the Latin as a guide:

• Πλάστης → Creator (twin col. 0063, Διὸ καὶ ὁ Πλάστης … ἑωρακώς = Hinc vero et Creator … cerneret). The anchor fixes Κτίστης = Creator and πλάστης = Fashioner. Following the Latin would have merged the two maker-words in the one sentence where the argument turns on moulding (τὴν τοῦ πλάσματος περιθέσθαι μορφήν).
• πάναγνος → innocentissima / purissima — Allatius uses at least three Latin words across this batch for the one Greek epithet.

Register decision fixed for this batch (flag for merge)

παράδοξος and θαυμαστός stand side by side in §21 (τῇ παραδόξῳ … προόδῳ … τοῖς ἔτι παραδοξότερον ἐσομένοις … θαυμαστὰ μὲν ᾖ τὰ κατ᾽ αὐτήν). Fixed here: παράδοξος = wondrous (anchor precedent, παραδοξότατον = "most wondrous"), θαυμαστός / θαυμάσιος = marvellous, θαῦμα = wonder, θαυμάζω = marvel. Sermon I's English used "wondrous" for both words in different chunks; within Sermon II they are kept apart. Merge session: check chunks 0000–0002 for the same pair and unify on this assignment — one voice within a work is required (the style file's boundary paragraph).

Likewise ποίησις = making where it is contrasted with a maker (0004 §27, ἥττω ποίησιν), and "creation" where it names the created whole (πᾶσα ἡ ποίησις); ἄβατα / ἄδυτα are kept apart as untrodden places / inner sanctuary throughout.

2. Scripture — every lemma read word by word against the printed Greek

Marked [var:] (one, in 0003):

1. 0003, before [0064] — Mk 10:14. Migne prints Ἄφετε τὰ παιδία ἔρχεσθαι πρός με, καὶ μὴ κωλύετε αὐτά· τῶν γὰρ τοιούτων ἐστὶν ἡ βασιλεία τῶν οὐρανῶν, and the twin's own footnote cites Marc. x, 14 — but Mark reads ἡ βασιλεία τοῦ Θεοῦ, and the bare μὴ κωλύετε without καί. Both printed features are Matthew's (19:14). Two-witness concurrence: the Latin column gives talium est enim regnum coelorum and et ne prohibueritis eos, so the conflation is the plate pair's, not our file's. Rendered as printed; the marker names Mark's reading.

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

• 0003 (opening) — Luke 2:52. προέκοπτε σοφίᾳ καὶ ἡλικίᾳ καὶ χάριτι παρὰ Θεῷ καὶ ἀνθρώποις, without the received ἐν τῇ, and applied to the Virgin, not to Jesus. The refit is the homilist's and is visible in his own frame (ἵνα καὶ τοῦτο προσθείην). Latin concurs (proficiebat sapientia et aetate et gratia apud Deum et homines).
• 0003 §22 — Ps 18:15 (LXX). ἡ μελέτη τῆς καρδίας αὐτῆς … ἐνώπιον Θεοῦ διαπαντός for the received τῆς καρδίας μου … ἐνώπιόν σου. Refitted to the third person and announced (κατὰ τὸν προφήτην φάναι Δαβίδ). Latin concurs (meditatio cordis ejus … in conspectu Dei semper); Migne cites Psal. xviii, 15 on the plate.
• 0003 §23 — Luke 12:42–44. Migne prints ὁ πιστὸς οἰκονόμος καὶ φρόνιμος (received: ὁ πιστὸς οἰκονόμος ὁ φρόνιμος) and τοῦ δοῦναι (received: τοῦ διδόναι). Latin concurs on both (fidelis dispensator et prudens; ut det) — two-witness concurrence, so the readings are the plate pair's. Rendered as printed. Note that Isidore has just glossed τὸ σιτομέτριον in his own voice, so the lemma is load-bearing: "measure of wheat" is kept identical in quotation and gloss.
• 0003 §23 — Ps 4:9 (LXX). Σὺ, Κύριε, καταμόνας ἐπ᾽ ἐλπίδι κατῴκισάς με, dropping the received opening ὅτι. Truncation, not a variant. Ballerini's note (71) on the plate already adjudicates the singulariter / καταμόνας question between Vulgate and LXX — the plate's own apparatus settles it, so no marker of ours (the Nicetas calibration datum).
• 0003 §23 — Ps 17:7 (LXX). Ἤκουσεν ἐκ ναοῦ ἁγίου αὐτοῦ φωνῆς μου· καὶ ἡ κραυγή μου ἐνώπιον αὐτοῦ εἰσελεύσεται εἰς τὰ ὦτα αὐτοῦ — verbatim LXX, no divergence. Checked because Ballerini's note (72) draws attention to the Vulgate's introivit against the Greek future; the Latin column here prints the future introibit, i.e. it follows the Greek and not the Vulgate. Worth recording as a counter-instance to the "Latin conforms scripture to the Vulgate" convention.
• 0003 §24 — Luke 1:49 woven, not quoted (ἡ τῶν μεγαλείων τυχοῦσα τοῦ δυνατοῦ); §25 — Luke 3:22 woven (καταβῆναι σωματικῷ εἴδει τὸ Πνεῦμα τὸ ἅγιον). Both cited by Migne on the plate.
• 0003 §26 — Ps 4:7 (LXX). ἐπ᾽ αὐτὴν … τὸ φῶς τοῦ προσώπου Κυρίου for the received ἐφ᾽ ἡμᾶς … τοῦ προσώπου σου, κύριε. Woven into the homilist's own ὥστε-clause and explicitly attributed (κατὰ τὸν Δαβίδ), so the refit is authorial. Latin concurs (signatum fuerat super eam lumen vultus Domini).
• 0004 §28 — Heb 6:19–20. ἣν ὡς ἄγκυραν ἔχομεν τῆς ψυχῆς … ἀσφαλῆ τε καὶ βεβαίαν, καὶ εἰσερχομένην εἰς τὸ ἐσώτερον τοῦ καταπετάσματος, ὅπου πρόδρομος ὑπὲρ ἡμῶν εἰσῆλθεν Ἰησοῦς — verbatim the received text, with the homilist's κατὰ τὸν μακάριον Παῦλον εἰπεῖν interposed. The one material change is the referent: in Hebrews the anchor is hope; here the antecedent of ἥν is the Virgin's grace (τῇ τῆς πανυμνήτου δὴ ταύτης χάριτι, ἣν ὡς ἄγκυραν ἔχομεν). The English keeps the relative bound to "grace" and does not slide to "hope." Latin concurs (gratiam ejusdem … quam … sicut anchoram habemus animae). No marker: the words are the received words, and the transfer is the sermon's own argument, stated in its own syntax.

3. The Latin-twin pass, per chunk (pilot §6)

Method as specified: draft first, then one deliberate pass against the twin hunting (1) Greek that made no sense, (2) different asserted facts, (3) polarity disagreement. The third witness was greppable throughout — both chunks have full leaf coverage (twin leaves 29–31, 32–33; scan-Greek the same leaves).

• Chunk 0003 — NOT clean. One [lat:] fired (§4), one [var:] (§2), numerous ladder-row-1 OCR corrections (§5). Polarity collation found no disagreement: ποῖον οὐ κάτω τίθησι θαῦμα; / quodnam aliud … non vicerit? · οὐκ αἰδεῖται τὴν ἧτταν / primas eisdem cedere haud [veretur] · οὐκ ἄν τις ἁμάρτοι καλέσας / neutiquam a vero … abluderet · οὐδὲν μὲν γὰρ ὑψηλότερον / Nihil enim … majus · οὐκ ἦν τι τῶν ἀνθρωπείων ἰδεῖν / nihil ex humanis in oculos incurreret · οὐδὲν ἀνάξιον ᾤετο / haud indignum duxit · βραχέα ἢ οὐδὲν … δεῖσθαι / vix aut ne vix quidem … indigeret · the double negative at §23 ἀλλ᾽ οὐκ ἂν ἔχοι τις … μὴ … δοῦναι συνεργεῖν / nequaquam tamen concedere quispiam poterit, nihil prorsus mansionem illam … profuisse — both columns carry two negatives and the English carries both ("one could not … grant that it did not work together"), not the smoothed positive · ἀθέατος οὐκ ἐπ᾽ ὀλίγον κατέστη / haud breve tempus invisibilis latebat — see the note below · οὔτε γὰρ … ἀνέῳγε τὰ ἄβατα / non vero cuipiam alteri · οὐκ ἀεὶ δὲ καὶ τοῦτο / neque vero id semper · ἃ μὴ βάσιμα τοῖς πολλοῖς / quo utique non ascendunt turbae · Ἀλλ᾽ οὐδέ γε … εἶχεν ὁ θεατής / neque … comprehendere spectator valebat · οὐ πᾶν ὅσον περὶ Θεὸν ἐνενόησεν / non ideo omnia … perviderit · οὐκ ἦν ἂν, ὅς … / nequaquam hic assequebatur · Θεῷ δ᾽ ἐξ ἴσης οὐδέποτε / absit tamen ut parem unquam eam Deo habueris. All carried on both sides.
• The one negation where the third witness is the weaker reading, and why Calfa was kept: at §24 Calfa prints ἀθέατος οὐκ ἐπ᾽ ὀλίγον κατέστη ("she became unseen for no short time"); the scan's Greek OCR at that point reads ἀθέατος íx' ὀλίγχον κατέστη, i.e. it has lost the negative in a run it is visibly mangling. The Latin decides it: ideo haud breve tempus invisibilis latebat. Calfa + Latin agree, the scan is damaged — the negation stands and is carried. Recorded because dropping it would have reversed the sentence with no lexical trace (Pattern 7a mechanism 2).
• Chunk 0004 — the twin pass found NOTHING under headings (2) and (3). No [lat:], no [var:], no polarity disagreement, no asserted-fact divergence, no numerals of any kind. Only ladder-row-1 OCR corrections (§5). Polarity collation, in full, since a null result is only checkable if it is itemized: οὐκ ἂν ἔλαττον ἢ νῦν φρονεῖν ἦκον ἄνθρωποι Θεοῦ / nihilo tamen minus quam modo homines … pervenire potuissent — the load-bearing counterfactual of §27, and the two columns agree · οὐδ᾽ ὅσον ὑποβέβηκεν / ne ille vilescunt, quantum nec cogitatione assequi fas est · οὐκ ἄρ᾽ ἀνάγκη … προσεῖναι / hic profecto non indiget · μηδενὸς τῶν γενητῶν ὑποστάντος ἑτέρου / etiamsi nulla alia e creatis rebus exstitisset · τῶν ὄντων οὐδέν, Θεοῦ δίχα, μεγαλοπρεπέστερον ταύτης / nihil eorum quae exsis[tunt], excepto Deo, est ipsa magnificentius · οὐ τοσοῦτον τοῦτο δοκεῖ θαυμαστόν, ὅσον εἰ μηδὲ τὴν ἀρχὴν τοῦτ᾽ ἐγεγόνει / non adeo stupendum hoc videtur, ac si ne id quidem factum fuisset — the sermon's sharpest polarity trap, since the sentence argues that the miracle would be more marvellous had it NOT happened; both columns print it so and the English carries it · οὐ γὰρ … θαυμάζειν χρή· ἀλλὰ τοὐναντίον / Neque enim est cur miremur … illud potius · Ἐπεὶ δὲ οὐδ᾽ ἄν … τὸν εἰκότα ἄν ποτε πλέξαιμεν κρότον / cum quidquid demum … nunquam parem meritis coronam laudum texere possimus. No disagreement anywhere.

Numerals (the named class). The batch contains no in-prose letter-numerals. The age statements are spelled out — τριετής (§21) and μέχρι τριετίαν (§24) — and both agree with the Latin (adhuc trimula, usque ad triennium). The section marks are handled in §6. No numeral [lat:], and none was near firing.

Reciprocal-crop check (pilot §4a). Both crops in this batch bleed the other column: the twin's 0003 carries a column of stray Greek (καιρῷ, … αὖ- ὑμῖν … ὁρατή) and the scan-Greek crop carries running Latin. Every apparent Latin gap in the twin was checked against the Greek crop before any conclusion was drawn from it — this is what confirmed the semel per annum addition of §4 is Allatius's and not a truncation artifact (the phrase survives complete in the twin's own line, semel per annum in Sanctasanctorum introeunte, and its second occurrence survives in the Greek crop's Latin bleed, anno flebat visibilis).

4. The one [lat:] — Allatius supplies the annual entry

FIRED — 0003 §25, Greek col. 0068 / Latin col. 0067.

Greek: Τὸ δέ γε μόνῳ τῷ ἱεράρχῃ, ἅπαξ εἰσιόντι τὰ ἄδυτα, ὁρᾶσθαι τὴν πάναγνον οὐκ ὀλίγον… Latin twin, col. 0067: Quod vero ab uno summo sacerdote, semel per annum in Sanctasanctorum introeunte, innocentissima conspiceretur…

The Greek says only once; the Latin says once a year. A reader of the Latin column takes away a fact about the Levitical calendar (Heb 9:7) that the Greek column does not assert. Greek plate-verified against the third witness: the scan's own Greek column reads ἅπαξ εἰσιόντι τὰ ἄδυτα with no temporal qualifier, independently of Calfa.

It is not a one-off slip on Allatius's part, which is what decided the marker: he does the same thing again eight lines later, where the Greek's καὶ ἅπαξ ἦν θεατή, καὶ μόνῳ τῷ ἱεράρχῃ is rendered …anno fiebat visibilis et a solo summo sacerdote videbatur (the phrase survives in the Greek crop's Latin bleed at leaf 31). Twice-repeated supply of a specific fact is past the boundary Pattern 16 draws around free rendering. One marker at the first site; the second is recorded here rather than marked, so the page is not annotated twice for one habit.

This also is why ἱεράρχης is fixed as hierarch and not "high priest": the Latin's summus sacerdos is the vehicle of the same importation.

Candidates considered and REJECTED (named so a later pass does not re-raise them) — all Allatius's normal freedom, not divergence:

• ἀνθρώπων μόνον τῶν ἱερῶν Θεὸν ὁρᾷν / ad Dei aspectum … electis tantummodo contingit hominibus — "the consecrated" vs "the elect". A different word, not a different fact.
• Θεὸς ἐπίγειος / divinitas erat terrestris — the abstract for the concrete.
• οὔτε γάρ … καὶ ὑπὲρ μόνου τοῦ Σωτῆρος restructured as Quemadmodum enim … sic — Allatius repairs a broken correlative (see §7); syntax, not fact.
• ἐπὶ τῆς θεραπείας αὐτοῦ (Luke 12:42) / super familiam suam — the Vulgate's word in a scripture quotation: convention, and the lemma-and-gloss rule is explicit that this never fires.
• The whole Ballerini apparatus in both crops (notes 68–76, the immaculist annotations) — plate apparatus, verifier-only, not ours to mark.

5. Ladder row 1 — Calfa OCR damage corrected silently, evidence cited

Per the ladder and Pattern 13a clause 1: words lost or mangled mid-sentence with a certain reading are restored silently in the English and logged here. Nothing in this batch was a reader-visible structural unit, so no [ed:] fired; nothing was plate-verified carried type, so no [sic:] fired.

The intrusive-Κ class continues — the standing Calfa artifact in this PG 139 file (seven attested sites before this batch, across sermo-1 and Nicetas). Four more here, all corrected silently: τὰς Κλλας παρθένους → τὰς ἄλλας (0003) · Θεῶ Κμόνῳ προσλαλοῦσα → Θεῷ μόνῳ (0003) · μείζονος Κἔργου → μείζονος ἔργου (0004) · and the same hand in ΚΜΗʹ for ΚΗʹ (§6). The class now stands at eleven sites in three works of this volume.

Chunk 0003, per site (✱ = confirmed against the scan's Greek column): σοφίμ → σοφίᾳ ✱ · ὁ δωτὴρ → ὁ Σωτὴρ ✱ (Lat. Salvator) · Ὦφετε → Ἄφετε ✱ · κωλόετε → κωλύετε ✱ · τν γὰρ τοιούτων → τῶν γὰρ τοιούτων ✱ · μορφωῦῆναι → μορφωθῆναι ✱ · ἢ μελέτη → ἡ μελέτη ✱ · ὁιιοίωσιν → ὁμοίωσιν ✱ · Λιὸ καὶ ὁ Πλάστης → Διὸ ✱ · ὑώηλοῦ → ὑψηλοῦ ✱ · κεκοινονηκέναι → κεκοινωνηκέναι ✱ · περιῆν εἰει ἀγνείαν → περιῆν εἰς ἁγνείαν ✱ · εὐχῆς χάοιν → χάριν ✱ · Τ δὲ σιτομέτριον → Τὸ δὲ ✱ · θαυμααίως → θαυμασίως (scan mangles the same run, σίως; Lat. excellentissime) · δοκῶμοι → δοκῶ μοι ✱ (split/run-together type) · Τς ἄρα ἐστὶν → Τίς ἄρα ✱ · φρόνιμος ν καταστήσει → φρόνιμος, ὃν ✱ · τῆς θεραπεί ας ιὺτοῦ → τῆς θεραπείας αὐτοῦ ✱ · ακάριος → μακάριος ✱ · εἰρήσει → εὑρήσει ✱ · Ἀληθώς λέγωὑμν ἄτι → Ἀληθῶς λέγω ὑμῖν ὅτι ✱ · ἐκείνο προεῖπε Δαβίδ → ἐκεῖνο ✱ · εἰσελεύσειαι → εἰσελεύσεται ✱ · κοινωνων ἠν → κοινωνῶν ἦν ✱ · κἀλλος → κάλλος ✱ · έδαφος → ἔδαφος ✱ · προῦπέγραψεν → προϋπέγραψεν ✱ · συν.ῖναι → συνεῖναι ✱ · σωματικῷ εἰδει → εἴδει ✱ · Λεσπότου → Δεσπότου ✱ · ἁνθρώπων μόνον → ἀνθρώπων ✱ · ἰεράρχῃ → ἱεράρχῃ ✱ · ἑωτεκὼς → ἑωρακὼς ✱ · ὡφι εἰπεῖν → ὡς εἰπεῖν ✱ · ἐσμειώθη → ἐσημειώθη ✱ · σχεθὸν → σχεδὸν ✱.

Two Calfa artifacts of a different kind, also corrected silently:

• A stray floating capital. Calfa prints τὸ φῶς τοῦ προσώπου Κιρίοι, Ο ὥστε σαφεῖς…; Κιρίοι is Κυρίου and the isolated Ο is line-noise — the scan's Greek reads τοῦ προσώπου Κυρίου, ὥστε σαφεῖς with nothing between. Dropped.
• A spurious mid-sentence full stop. Calfa prints …διατριβῆς ἐκείνης. μὴ πρὸς θείαν ὕψωσιν δοῦναι…, which severs the double negative that carries the clause. The scan's Greek has no stop there and the Latin runs on (mansionem illam … profuisse). Removed. Pattern 8 was checked before doing this and does not apply: Pattern 8 protects punctuation Migne prints, and the third witness shows Migne does not print this one.

Chunk 0004, per site (✱ = scan-confirmed): ποῖησις → ποίησις ✱ · ἦλιος → ἥλιος ✱ · ταῦτα λὲ μὴ → ταῦτα δὲ μὴ ✱ · τῆς ἀπαραμίλλου μόης → μόνης ✱ · ψποβέβηκεν → ὑποβέβηκεν ✱ · ουγχωρήσειεν → συγχωρήσειεν (scan has the same defect, ουγχωρήσειε — see §7) · Ἢρκει → Ἤρκει ✱ · τοῦ Θεου → τοῦ Θεοῦ ✱ · τα λοιπὰ → τὰ λοιπὰ ✱ · εῖη → εἴη ✱ · ούτη φανείη → αὕτη ✱ · οἱς ἂν → οἷς ἂν ✱ · διὰ τῆς Παμθένου → Παρθένου ✱ · Ὃθεν → Ὅθεν ✱ · λέ, ω δὲ → λέγω δὲ ✱ · τῆς τανάγνου → τῆς πανάγνου ✱ · δυκεῖ → δοκεῖ ✱ · τοὺναντίον → τοὐναντίον ✱ · τῶ ἰσχηρῷ → τῷ ἰσχυρῷ ✱ · Εἰη δὲ ἡμὶν → Εἴη δὲ ἡμῖν ✱ · ὅγκυραν ἐχομεν → ἄγκυραν ἔχομεν ✱ · τἰσερχομένην → εἰσερχομένην ✱ · καταπετάσμμτος → καταπετάσματος ✱ · Ἰὴσοῦς → Ἰησοῦς ✱ · ἀγίῳ Πνεύματι → ἁγίῳ ✱ · αἰωνων Ἀμήν. → αἰώνων. Ἀμήν. ✱ — the terminal stop before Ἀμήν is supplied: Calfa has none, the scan's Greek prints αἰώνων ᾽Αμήν. and the Latin saeculorum. Amen. This is the sermon's last mark and it is a digitization loss, not a plate reading; restored silently per 13a clause 1, recorded here because it is the closing punctuation of the work.

6. Section numerals — both are Calfa damage, both third-witness verified

Rendered as arabic + period per the anchor. Neither is a [lat:]: in both cases the scan's own Greek column agrees with the Latin against Calfa, which the ladder's §4a makes decisive — the defect is certainly ours.

7. Construal cruces — printed reading kept, conjecture here

• 0003 §20 (opening) — περὶ οὗ μόνου, masculine, of the Virgin. Calfa prints the masculine/neuter relative where the referent is the Virgin, and the Latin has the feminine in ipsam solam. The scan's Greek has lost the relative at a line start, so it cannot arbitrate. No action and no marker: Pattern 9 — Greek relative gender has no English exponent ("concerning whom alone"), so there is nothing to emend and nothing to conceal. Logged so a later sweep does not churn it.
• 0003 §23 — ἕως οὗ καὶ τῇ γῇ παρὼν ὁ Σωτὴρ … δείκνυται χαίρων. ἕως οὗ + present indicative is temporal ("so long as"), but the clause is plainly causal in force, and Allatius reads it causally: quando ipse quoque Salvator, in terris cum versaretur, recessu et monte … delectatus cernitur. Two-witness concurrence on the Greek — the scan prints ἕως οὗ too — so the difficulty is the plate's, not our file's. Rendered literally, "so long as," rather than smoothed to "seeing that." This is exactly the place a silent repair would have been invisible: the causal English reads better and leaves no trace. The Latin's construal is recorded, not adopted.
• 0003 §22 — τούτῳ χαρίσηται τὸ κατελθεῖν, τῆς τε εἰκόνος αὐτῷ κεκοινωνηκέναι. The dative αὐτῷ is ambiguous between "with him" (the Fashioner) and a resumptive of the recipient; the Latin binds it the other way (eamdem cum ea formam … communia cum ipsa haberet, "with her"). The English keeps the Greek's own dative — "the having shared with him both his image and the rest of his manner of life" — and the Latin's binding is recorded rather than silently taken, per the standing rule that the Latin may never settle a reading without a crux saying so.
• 0003 §25 — the broken correlative οὔτε γάρ … καί. οὔτε γὰρ, ὡς τῇ Παρθένῳ, ἑτέρῳ τῳ … ἀνέῳγε τὰ ἄβατα, καὶ ὑπὲρ μόνου τοῦ Σωτῆρος … προὔθηκεν ἑαυτόν — one expects οὔτε … οὔτε, or ὥσπερ … οὕτως. The scan's Greek prints the same broken correlative, so it is the plate's. Allatius repairs it (Quemadmodum enim … sic). Rendered with the break kept ("For neither … and …"), under Pattern 7; the same treatment sermo-1 gave the Rubim οὔτε … καί at col. 0028. Not a [lat:]: syntax smoothing, not a different fact.
• 0004 §27 — ἵν᾽ ὅπως εἴη … κρείττων αὕτη φανείη. Calfa's doubled purpose/manner particles are awkward and the scan's Greek prints only ἵν᾽ εἴη; the Latin has ut apparere posset, quot et quantis et quam magno intervallo ipsa … excelleret. Because the scan's loss here is a plain line-edge drop rather than a competing reading, Calfa's fuller text is kept and rendered "that it might appear in what manner, and by how much, she is better than things so many and so great." Flagged as the one place in 0004 where the two Greek witnesses differ in extent.
• 0004 §27 — ουγχωρήσειεν. Both Calfa and the scan's own Greek column print the same non-word for συγχωρήσειεν (scan: ουγχωρήσειε). Since the identical defect appears in two independent OCRs of the same plate, this is more likely a shared optical failure on ου/συ than a plate reading; corrected silently rather than carried as [sic:], because a [sic:] is a public claim about Migne's type and the two OCRs are not two readings of it here. If a leaf render ever shows the plate genuinely printing ουγχωρήσειεν, this is the site that turns into a [sic:].

8. Ratio, shape, and where the work ends

Greek 1015 → English 1482 (1.46×); Greek 392 → English 573 (1.46×). Both track the anchor's homily ratio of 1.49×, well clear of the chronicle 1.25×.

Column anchors: 0003 carries [0064] [0065] [0068], 0004 carries [0069] [0072] — verbatim, in order, 1:1 with the Greek chunks (verified by count). No ## heads, no italics and no list lines exist in either Greek chunk, so none were introduced.

Where the sermon ends. Chunk 0004 is the work's last: the doxology ᾧ τό τε κράτος καὶ ἡ προσκύνησις σὺν τῷ Πατρὶ καὶ τῷ [0072] ἁγίῳ Πνεύματι εἰς ἀτελευτήτους αἰῶνας τῶν αἰώνων. Ἀμήν. closes in column 0072, and the [0072] anchor falls inside the doxology, between "the Father and the" and "Holy Spirit" — reproduced there, mid-phrase, as printed. The doxology's formula (εἰς ἀτελευτήτους αἰῶνας τῶν αἰώνων. Ἀμήν.) is identical to Sermon I's and is rendered identically ("unto the unending ages of ages. Amen."); the governing relative differs (Sermon I doxologizes the Trinity, ᾗ πρέπει δόξα; Sermon II doxologizes Jesus, ᾧ τό τε κράτος καὶ ἡ προσκύνησις σὺν τῷ Πατρὶ καὶ τῷ ἁγίῳ Πνεύματι), and the English keeps that difference visible.

9. For the merge session

1. παράδοξος / θαυμαστός — this batch fixes "wondrous" / "marvellous" respectively (§1). Chunks 0000–0002 must be checked and unified; Sermon I's English used "wondrous" for both.
2. πάνσεμνος = all-august, μεγαλουργική = great-working, ἱεράρχης = hierarch — new to the set's table; adopt into the work-wide conventions section when the two cruces files are merged, and carry them to Sermons III–IV.
3. ἄβατα = untrodden places / ἄδυτα = inner sanctuary — kept apart here; confirm 0000–0002 does the same, since the sermon's whole subject is the entry into them.
4. The [lat:] at §25 is the batch's only marker of its kind; the second occurrence of the same Allatius habit is deliberately unmarked (§4). If the merge prefers both marked, the site is καὶ ἅπαξ ἦν θεατή.
5. Frontmatter language — kept "grc" per the shipped PG convention; do not flip it for this work alone.