Epistola ad Hebraeos

PL 114, cols 0643A–0670D · 13 chunks · 11,438 Latin words. Attributed by Migne to Walafrid Strabo; in fact Anselmus Laudunensis et schola. Translated 2026-08-30 by three Opus stints, none of which could see the others.

This file is the MERGE LAYER, and it is not a copy of the range files. cruces-0000.md (chunks 0000–0004, Heb 1–4), cruces-0005.md (chunks 0005–0008, Heb 5–9) and cruces-0009.md (chunks 0009–0012, Heb 10–13) stand beside it and carry the per-site detail — the full lemma collations, the negation tables, the unexpressed-subject rulings, the plate candidates. This file records only what was settled across the three ranges, or after them by the plate: the thing no stint could do for itself.

⚠ Kept to the merge layer alone, per the 8986 finding: build-cruces.mjs concatenates every cruces*.md in the directory, so a merge that embeds its range files publishes the apparatus twice.

§0 · Apparatus, counted from the English after the plate gate

14 [var:] · 4 [cj:] · 3 [sic:] · 1 [d:] · 0 [ed:] · 0 [nt:] · 0 [f:] — counted from src/english/8993/0*.md after the plate gate and the polarity repairs, not summed from the three stints' reports.

The three stints filed 14 [var:] · 5 [cj:] · 1 [d:] · 0 [sic:]. The plate gate withdrew two [cj:] (§2) and added three [sic:] that no stint could have fired, because firing them required the page (§3); the polarity read added the fourth [cj:], on the elided negation at 0655C (§5).

verify-english.mjs 8993 → OK, 13 chunks, columns/notes/sections aligned. scan-raw-markers.mjs → no marker rendered as literal bracket text. ✅ plate-gate.mjs 8993 PASSES.

⭐ All three stints independently declined to fire a single [sic:], each giving the same reason: a [sic:] accuses Migne's type, our Latin is Corpus Corporum's transcription and not the plate, and no column had been read. That is now the third work running where the stated marker policy produced zero false accusations — and this time the plate proved the instinct right in detail, since four of the non-words they refused to accuse Migne of turned out to be Corpus Corporum's own (§2).

§1 · THE PLATE GATE — 14 pages, and what it returned

Pages 327–340 of raw/scans/pl114/patrologiaecurs06saingoog.pdf, read at 150 dpi full width with 400 dpi crops for every letter-level call, by the session directly rather than by fenced agents — deliberately, per the 8986 finding that briefing several readers on one suspicion produces convergence rather than corroboration. Map PDF page = (column + 11) / 2, calibrated at the printed corner numbers (p. 327 prints 643/644) before the first read.

14 [var:] · 14 STAND · 0 REVISED · 0 FALSE · 0 UNRESOLVED.

Every one is a multi-word scriptural divergence — the class a single-letter CC misread cannot manufacture — which is the same policy that returned 36/37 on 9006 and 7/7 on 8986. The divergences confirmed at the plate include Migne's Paulo minus minoratus est ab angelis (0644B), Ibunt hi in ignem aeternum (0645C), erit omnia in omnibus (0647D), non erubescit fratres vocare (0647D), Offensus (0649B, where Migne then prints infensus himself two lines on), fixit Deus (0656C), Tabernaculum enim sanctum (0658A), Aspersi corde (0661D), Fide qui vivit (0664A), Vis nos interficere (0665B), Si vos persecuti fuerint in una civitate (0665C), Habeamus gratiam (0669A) and Confidentium (0670A).

The [d:] at 0661A stands: the plate prints ad implendam tuam voluntatem tuam.

The three surviving [cj:] stand as Migne's own printing — Confessus twice (0644B, 0646A, where consessus is what the sense of the session at the Father's right hand requires) and ligna for signa (0646D). Each is additive under Pattern 18: Migne's word keeps its English.

§2 · ⭐⭐ TWO [cj:] WITHDRAWN — the conjecture was right and the CLASS was wrong

Both stints that fired a letter-level [cj:] were right about the reading and wrong about whose error it was, and the plate says so in both directions:

⛔ This is the class to carry forward. A [cj:] is milder than a [sic:] — it accuses nobody, it only offers a reading — but it still presupposes that Migne printed the word it conjectures against. Where the printed word is actually our transcription's error, the right outcome is a TEI patch and no marker at all, and the reader meets clean Latin instead of an apparatus note about a word Migne never set. Both sites read perfectly as conjectures and could have shipped that way forever.

⚑ verify-english.mjs caught both the moment the patched Latin landed — "a conjecture may only be offered against a word Migne actually prints" — which is the check doing exactly the job it was written for, without anyone pointing it at these two.

§3 · NINE HIDDEN DEFECTS, and the two that were MIGNE'S

Full list and evidence: data/tei-patches/8993.json. Two directions, as in 9006:

NORMALIZED — Migne's own defect, silently tidied by Corpus Corporum. Our Latin read CORRECTLY, so no check we own could have found either. Each is now restored and carries a [sic:]:

• 0650A — Et requievit dia septima. Migne prints dia, a non-word; he sets Septima die in roman two lines below on the same plate, so the two letters are comparable in his own type. Found while reading the column for a different marker.
• 0670B — Sicut generaliter ominibus scribit, ita generaliter omnibus praelatis. The defective sort and the correct one on one line, which is what makes it certain rather than a reading of worn type. Found while reading the line for propositos, four words away — addendum C earning its place a fifth time.

A THIRD defect of Migne's needed no patch — our text already agreed with the plate: 0669D serviendium for serviendum. The stint had carried it untranslated pending a read; it now takes its [sic:] and its English, per test 1 (a marker must not swallow the word it marks).

CORRUPTED — Corpus Corporum broke what Migne set correctly. Four letter-level (0665A padebat→pudebat, 0667B sanctur→sanetur, plus the two withdrawn-marker sites in §2) and two line-break failures (0654A, 0655D). ⚠ 0654A carried two faults in one phrase: CC failed to rejoin genu | erunt and inserted an in the plate does not print (qui eos in fide genuerunt). Nobody could have seen the insertion.

⭐ The two line-break sites are not the same fault, and the difference is worth recording: 0655D prints Christus per- with a hyphen that CC dropped; 0654A prints genu at the line end with no hyphen at all, so there Migne set none. Same repair, different cause.

§4 · ⛔ ADDENDUM A DOES NOT HOLD IN THIS WORK — non-words split THREE ways

The 2026-08-23 addendum records 10 for 10: every non-word suspicion confirmed as our corruption, every real-word suspicion refuted. The real-word half held perfectly here — absumptam (0652B), securint (0663A) and accipiunt sacrificium (0655A) were each suspected, each argued well, and each proved to be exactly what Migne prints. The non-word half did not. This work's non-words came back:

• CC's corruption — sanctur, padebat, vetata, Ac (4)
• CC's line-break failure — genu erunt, per manet (2)
• ⭐ MIGNE'S OWN — serviendium (1), and two more that were invisible as non-words because CC had already tidied them away (dia, ominibus)

⚑ So the rule to state is the weaker one the addendum's own evidence supports: a non-word is worth raising as a plate candidate, and raising it is right every time. What it is evidence of — our transcription or Migne's type — the frequency argument cannot tell you, and this work is the counter-example that proves it. The ship test is unaffected and did its whole job: every one of these was carried with a crux and no marker, so nothing false went out while they were unresolved.

⭐⭐ The strongest single result of the run is 0667B sanctur, where the stint predicted the plate without seeing it. It grepped the corpus, found the same e→c fault at a second site in another work (11065 @0044, vel sanctur infirma sitting beside its own bene sanatur), reasoned that one fault shared across two works indicts the transcription rather than the type, and on that ground withheld a [sic:] here and on three other non-words. The plate prints sanetur. Four false public accusations against Migne, prevented by an argument made from the files alone.

§5 · THE POLARITY READ — three blind readers, 13 sites raised, 4 acted on

sites: 4 · ours: 4, recorded via polarity-record.mjs. Every reader named its clean chunks.

⭐ The one serious finding, and it is the class that leaves no trace — 0655C VERS. 16. Migne prints Non secundum legem. Id est secundum carnalem intellectum legis… — the gloss leans on the lemma's non, and prints no second one. Our English put a full stop after the lemma and then read as a positive assertion: that he acted according to the carnal understanding. The reader proved the host structurally from Heb 7:16's own non … sed, and proved the house practice from this same work, which carries an elided lemma word into the gloss at 0660A. Resolved per the settled ruling (8954 @0593C; precedent 7561 @0648): the text is NOT touched — 7a forbids supplying a negative — and a [cj:] names the elision. The marker names what Migne elided, not a fault of his.

The other three, all ours, all repaired: 0655C VERS. 14, where eadem/facta est are feminine and the subject is the tribe, not the priesthood — our bare "it" made the sentence contradict its own previous clause; 0656C VERS. 4, where "he would not be a priest" attached non esset sacerdos to the hypothetical earthly man, and the sentence is impersonal — "there would be no priest"; and 0670A VERS. 12, where in quorum nullo is plural, so sensus is "the senses of the body" and our singular left the relative with no antecedent.

⛔ Four of the readers' findings were REFUTED by the plate, and three of those were their most confident. absumptam (0652B), accipiunt sacrificium (0655A) and id est non olim (0653B) are all exactly what Migne prints, as is pro illis at 0644C, which a reader flagged as a prae/pro letter-slip. ⭐ Every one of the four was an argument from what the sentence ought to say. The stints' rule — a real word is Migne's until a plate says otherwise — beat all four.

Declined, with the reason recorded so they are not re-raised: the ibi at 0665B (the referent is Egypt, from Heb 11:27, but Migne's own ibi is equally ambiguous, so the literal English mirrors it); the four unexpressed "he"s at 0656D (the subject is Moses, from Heb 8:5, and the Latin names him no more than we do); and the lemma at 0669B, where quidam latuerunt is Migne's own calque of Heb 13:2 and both readings construe.

⭐ 0649D — the dittography, settled at the plate and left standing. Migne prints omnes audierunt, sed non omnes audierunt, sed non omnes pervenerunt — vacuous, and his. The English renders it literally with no marker: not [d:] (the run is not a doubled setting), not [sic:] (all real words in real slots), not [cj:] — because obedierunt (Num 14:22's idiom) and crediderunt (Heb 4:2's argument, which the very next chapter makes) are both available and nothing in the file or on the page chooses between them. Naming a rival is not adjudicating between them; where neither wins, the crux is the honest place for both.

§6 · WORK-WIDE CONVENTIONS RECONCILED ACROSS THE THREE STINTS

So that a later sweep does not churn them:

1. Heads follow Migne's own FORM, not a single house form. CAPUT PRIMUM. → CHAPTER ONE. (spelled, because he spells it); CAPUT II.–CAPUT XIII. → CHAPTER II.–XIII. (Roman, because he sets Roman). The third stint reasoned this out independently and correctly: the work's own precedent governs, not Ruth's.
2. Vocabulary held across all three ranges with no drift: pontifex → "high priest", testamentum → "testament" (never "covenant"), sacrificium → "sacrifice", hostia → "victim", figura → "figure". ⚠ Note the one collision, which is the English language's and not a drift: hostia and victima both render "victim" (0005 has the only victimam). No passage contrasts the two, so nothing is lost.
3. VERS. n.-- passes through verbatim — 218 occurrences, no "VERSE" anywhere.
4. ⭐ The one real drift, and it is REGISTER: the third stint archaized the gloss's own voice with "ye" where the first two used "you". Twelve sites converted to "you" at the merge. Pattern 17's justification is what decides it: the archaism earns its place because it recovers number, a distinction the source marks and modern English cannot — and "ye" versus "you" recovers case, which English does not need, so it is archaism without the warrant. Pattern 17's own boundary is explicit that the scripture-echo licence "never licenses archaism in the author's own voice." Five "ye" remain, all inside quoted scripture — the lemmata For whereas ye ought (0653A) and Call ye to mind (0663A), and the Chrysostom and Augustine quotations at 0663A/0663B/0670A — where the licence does cover the quoted words. The sibling epistles corroborate the house voice: 8986 runs 2 "ye" against 100 "you".
5. No [nt:] anywhere — all 27 notes in the work are bare citations or sigla, not editorial prose, so Pattern 15 is not triggered. No [f:] — runbook false positive 4: a lemma-and-gloss commentary is not a florilegium.
6. deservit/deserit, the standing check, is discharged for the whole work: zero occurrences in chunks 0000–0004, clean in 0005–0008 (deserant only, correct), five instances checked correct in 0009–0012.

§7 · WHAT REMAINS OPEN

• 0661C — unsettled because our copy clips it. quod non valet ad probationem ___: the word our file reads as nostram falls inside p. 336's left-margin clip. It changes nothing in the English. The second witness (raw/scans/pl202/copyB_patrologiaecurs114unkngoog.pdf) would close it in one page.
• 0655D — Ilic autem or Hic autem? Raised by this read, by no agent. The plate's italic capital may simply have lost its crossbar at this scan's resolution. Not patched and not marked: it is precisely the worn-sort-versus-real-defect ambiguity that 9006's reader flagged as its least certain class, and the honest record is that it was seen and left.
• 11065 @0044 carries the same sanctur fault and is a shipped work. Per the no-retrofit rule (Wilson, 2026-08-18) this is recorded, not swept.

Agent 1 of three. Chunks 0000–0004, ARGUMENTUM + CAPUT PRIMUM–IV, Heb 1:1–4:16. This file is one third of the work's apparatus and must be merged with cruces-0005.md and cruces-0010.md (or whatever the peer ranges are named) before the work ships.

verify-english.mjs clean on all five chunks (the only failures it reports are peer chunks not yet written). Mechanical state of my five:

⚑ The band-by-band pass was not cosmetic: it returned fourteen sites where my first draft had softened or supplied a mark, every one of them invisible to a total-count check. Two were in the same band and would have cancelled. The list is at §5.

Apparatus fired: 3 [cj:] · 5 [var:] · 0 [sic:] · 0 [d:] · 0 [ed:] · 0 [nt:].

§1 · THE LEMMA WALK (7a″) — every span in cols 0643B–0651D, checked in place

Collated against sources/vulgate/clementine-flat.txt. The launch pre-scan flagged 21 spans in my range; 5 of those fall in chunk 0005 or later and are the next agent's. Of the 16 in my files, none was a substantive divergence — every one is an artefact of how this glossator cuts and reassembles his lemmata (§1.2). The five real divergences I found were mostly in spans the pre-scan marked ✓, which is the pre-scan working exactly as its own header says it does.

1.1 · The five divergences that took a [var:]

⚠ On the last one I considered and rejected the "Migne's own ( sic )" analogy (runbook false positive 5). It does not apply: that rule covers a defect Migne has flagged as a defect. Here he flags nothing — he prints two readings side by side as alternatives, which is a different act, and a reader still cannot tell from the page which one the Vulgate carries.

1.2 · The class that produced most of the ⚠ marks — non-contiguous reassembly

⭐ Name it before it convicts anything (the brief's closing rule). This glossator routinely prints as one lemma two or more words of the verse that the Vulgate separates, because he is citing the words he means to gloss, not quoting a continuous run. That is a citing practice, not a textual variant, and it is why so many spans failed a mechanical phrase search:

No marker on any of them, and the reason is Pattern 14's own gate: nothing is added, nothing is substituted, and no sense changes — only the order and the intervening words are the glossator's. Marking these would fire six [var:] that assert a divergence Migne never made.

1.3 · Real divergences I checked and declined to mark, with the reason for each

These are genuine differences from the Clementine. All are logged; none fired, and the ground in every case is Pattern 14's materiality gate ("in a way that changes the sense"), not doubt about the fact.

• 0643B Multifarie. — Clementine Heb 1:1 has Multifariam; multifarie occurs zero times in the whole Clementine (checked). It is nonetheless a well-attested Late-Latin form of the same adverb, in the same slot, with the same meaning; the gloss (multotiens) turns on the sense, which is unchanged. Rendered "In divers manners". Crux, no marker.
• 0643C Qui videt me, videt et Patrem meum — Jo 14:9 has videt et Patrem, without meum. An added possessive; the gloss's point (aequalis, quia non dissimilis) does not touch it. Crux, no marker.**
• 0646A Nonne sunt omnes, — Heb 1:14 has Nonne omnes sunt. Word order only. ⚠ I weighed the 7914 @1079B precedent (cinis et pulvis → "dust and ashes", where the gloss turned on the order) and it does not transfer: nothing here is expounded from the order. Crux, no marker.
• 0646A Et Spiritus distributionibus. — Heb 2:4 has et Spiritus Sancti distributionibus. An internal elision, not a terminal cut, so it is not the innocent "fragment" class; but the gloss expounds affluentiam gratiarum and no argument rests on Sancti. Crux, no marker.
• 0647A In eo autem. — Heb 2:8 has In eo enim. ⭐ The control that decided it is inside the work: Migne's own lemma at 0648B is In eo enim (Heb 2:18), so this book prints both connectives and neither is expounded. Crux, no marker.
• 0649C si initium substantiae, — Heb 3:14 has si tamen initium substantiae ejus. ⭐ tamen is not lost: it stands in Migne's own gloss frame three words earlier, hac tamen conditione. This is lemma-splicing, not omission. Crux, no marker.
• 0649D ex his quae audierunt. — Heb 4:2 has ex iis quae audierunt. Orthographic variant of the same pronoun. Crux, no marker.
• 0650A Et requievit die septima. — Heb 4:4 has Et requievit Deus die septima ab omnibus operibus suis. The subject is elided; it governs the unexpressed subject of the gloss that follows (§3). Crux, no marker.
• 0647D quando ei omnia subjicientur, et erit omnia in omnibus — this one did fire, but not as a straight variant: see §1.1's neighbour below. Migne's span opens with his own quando, so it is a paraphrase frame, not a quotation; but it drops the subject Deus that the Clementine supplies at I Cor 15:28, and ei two words earlier is Christ. The [var:] records the received reading precisely because the unexpressed subject is now ambiguous where the Vulgate's is not. Marker fired.

1.4 · Spans checked and found to agree with the Clementine exactly

Recorded because a list of findings alone cannot be told from a report by an agent that never looked. Every one of these was read word by word against the flat file:

Heb 1 (0643B–0646A): Olim · Deus · Loquens / loquens · Patribus · In prophetis · Multisque modis · In Filio · Constituit · Haeredem universorum · Qui cum sit splendor gloriae · Figura substantiae · Portansque omnia · Purgationem · Sedet · In excelsis · Tanto melior · effectus · prae illis · Hodie / hodie · Ego ero illi · Iterum introducit · Et adorent (both) · Angelos · Ad Filium · In saeculum · Unxit te · Deus · Et tu · In principio · Manuum · Peribunt · Tu autem permanebis · Veterascent · Et velut amictum · Mutabis eos · Tu autem · Sede · Pedum. Also Apc 21:1 at 0645D: Vidi coelum novum et terram novam agrees but for the dropped opening Et and the coelum/caelum orthography our source normalizes throughout. No marker.

Heb 2 (0646A–0648C): Propterea · Abundantius · Pereffluamus · Et omnis praevaricatio · omnis · retributionem · mercedis · Quid est homo · Visitas eum · Gloria · honore · Super opera manuum · subjecit · omnia, nihil dimisit · Eum autem · Propter passionem · Auctorem · Salutis eorum · Multos · filios · auctorem salutis eorum · Per passionem · Consummare · Qui enim · Ex uno omnes · Quia ergo · et ipse · Participavit eisdem · eisdem · Per mortem · Nusquam · Apprehendit / apprehendit · Per omnia · In eo enim · Qui tentantur. Also Jo 14:28 at 0647B: Pater major me est — exact. ⚑ assumpsit at 0648B is italic but is not a lemma: Migne prints it as the word the Apostle did not use (non autem assumpsit). Rendered as such.

Heb 3 (0648C–0649D): Unde fratres · Qui fidelis · Amplioris · Omnis namque domus · Qui autem omnia creavit · Deus est · Et Moyses · Quapropter · Hodie · Nolite obdurare corda vestra · Probaverunt · Infensus · proximus · In ira mea · Si introibunt in requiem meam · Cor malum incredulitatis · Donec hodie · Fallacia peccati · Participes enim Christi · usque ad finem, firmum retineamus · Quidam enim audientes.

Heb 4 (0649D–0651D): Non profuit · Quae audierunt · fidei · Et quidem · Operibus · hodie (×3) · quidem · Perfectis · Et in isto · Diem quemdam hodie · Post tantum temporis · Eis · requiem praestitisset nunquam · posthac · Sabbatismus · exemplum · Sermo Dei · Vivus · Pertingens · Divisionem animae · Ac spiritus · Compagum quoque · Medullarum · Habentes ergo · Non enim habemus · Qui non possit · In auxilio opportuno.

⚑ 0650C requiem praestitisset nunquam is a false ⚠. The pre-scan missed it only on the nunquam/numquam spelling and the Vulgate's comma: Heb 4:8 reads si eis Jesus requiem praestitisset, numquam de alia loqueretur, so Migne's three words are contiguous. See §4 for the scope question that sits on top of it, which is a real one.

1.5 · Two usque ad occurrences, decided per occurrence (addendum D)

• 0647D Probat, quod pro nostra liberatione Christum pati decuerit, etc., usque ad cum et utramque naturam per se constat scivisse. — the abridgment formula. Rendered "as far as", and no verb is supplied to complete the resumption tail. The etc. before it is the tell.
• 0649C …usque ad finem, firmum retineamus. — the verse's own words (Heb 3:14), and inside the italic lemma. Rendered as text: "we keep firm unto the end".
• ⚑ A third, at 0645A cum ostendit intro, usque ad corda hominum — the ordinary preposition, "even unto the hearts of men". This is the look-alike addendum D warns about, and it sits in the middle of an etymological gloss (intro + ducendum prised out of introducit), which is exactly where a formula-reading would have swallowed the argument.

§2 · THE THREE [cj:] — and the second-best reading in each case

2.1 · Confessus / confessus for consessus — 0644B and 0646A, TWICE

0644B Sic et per dexteram, non Deum deformavit, sed similitudinem honoris demonstravit. Confessus enim nihil demonstrat aliud nisi honoris aequalitatem. 0646A VERS. 13.-- … Sede. Victori Filio confessus offertur a Patre.

Both sentences are about Christ's session at the Father's right hand — the first follows per dexteram in the same breath, the second glosses the lemma Sede. Both slots demand a fourth-declension noun as subject (offertur; demonstrat), and confessus supplies none: it is a deponent perfect participle, so what the page says is "he that hath confessed", which is not what either sentence is arguing. Consessus, -us ("a sitting together, a session") fills both slots and both arguments exactly.

Fired [cj:], and Pattern 18's form kept: Migne's word first, in italics, then the conjecture. The English rendering stands beside the marker in both places and is not swallowed by it (test 1).

⚑ Why [cj:] and not [sic:] or a bare crux. Addendum A is the governing rule and it cuts both ways here: confessus is a real, well-formed word, so a corpus-frequency argument cannot convict it and a [sic:] would be a public claim about a plate I have not read. A [cj:] claims nothing about the plate — it says Migne prints confessus and proposes a reading beside it, additively. That is the marker built for exactly this case (11081's munde/mundo), and the bar it must clear is Pattern 18's: does the faithful English assert something the author did not? It does. "He that hath confessed shows nothing else but equality of honour" is a claim about confession where the argument is about the session.

⭐ The rival reading, and why it lost. The alternative is that confessus is Migne's and means what it says — confessio is genuinely a Hebrews keyword (Heb 3:1 pontificem confessionis nostrae; Heb 4:14 teneamus confessionem), and this glossator is reading Hebrews. That is not nothing, and it is why I record it. It loses on the syntax, not the theme: at 0646A the word is the nominative subject of offertur, and no participle of confiteor can be the thing "offered by the Father to the victorious Son". A thematic argument cannot rescue a slot a noun has to fill. ⚠ Wanted on plate: both columns, 0644B and 0646A. Two occurrences of the same substitution in one book is either a compositor's habit or our digitization's, and only the plate separates those (brief §4 — the normalization class runs in both directions).

2.2 · ligna for signa — 0646D

…neque tanta signa, neque tam diversa prodigia [0646D] per quod ostendit, quia non simpliciter est eis creditum, sed per ligna et prodigia.

Signa stands nine words earlier in the same sentence, paired with the same prodigia, and Heb 2:4's own signis et portentis is what the whole gloss is expounding. Ligna ("logs, timbers") is a real word and is rendered — "through logs" — with the conjecture beside it. The English asserts something false without the marker, which is Pattern 18's bar. ⭐ The rival I rejected: a [sic:]. It fails on the ship test — ligna is a real, well-formed word, addendum A is 4-for-4 that a corpus argument against such a word is refuted by the plate, and a false [sic:] publicly blames Migne for what may be Corpus Corporum's error. The [cj:] says the same thing without the accusation.

§3 · UNEXPRESSED SUBJECTS — every "he/she/it/they" I supplied, and what fixed it

The dominant defect class (12 of 25 sites on 11613). Named here so a blind reader can check my identifications rather than my prose:

• 0643C Non enim jam portio Domini est Jacob — no supplied pronoun; Jacob is the subject.
• 0644A Supra dictum est, quia per ipsum fecit omnia — "he made all things" = the Father, per ipsum = the Son. Fixed by Heb 1:2 per quem fecit et saecula, the verse under gloss two lemmata above.
• 0644A Postquam ostendit ejus excellentiam quo cuncta creavit et gubernat — subject of creavit/gubernat = the Father; quo = the Son ("of him by whom"). ⚠ quo is masculine/neuter and excellentiam is feminine, so quo cannot refer to the excellence; the antecedent has to be ejus, the Son. This is a Pattern 9 gender item, not a defect — no English exponent, no marker, and the reading it forces is the one I took.
• 0645A ubi jubetur adorari — the Son is commanded to be adored (Heb 1:6 Et adorent eum omnes angeli Dei).
• 0645D Significat mutationem coeli — subject = the lemma Et velut amictum, i.e. the verse. Rendered "It signifies".
• 0646D quia non simpliciter est eis creditum — eis = the preachers, not the ancients. Fixed by the very next sentence, cum credimus, non illis, sed Deo nos credere declaratur, where illis is unmistakably the preachers.
• 0647A Et ideo illius est memor — subject = God (from Cujus fuit memor Deus five lines above); illius = the earthly man.
• 0647D et erit omnia in omnibus — subject unexpressed. ei two words earlier is Christ, so the page says Christ; the Clementine's I Cor 15:28 says Deus. I rendered "he shall be all in all" (Migne's, undecided) and put the Vulgate's subject in the [var:] rather than choosing silently.
• 0650A Sex diebus fecit omnia — God, from the lemma Et requievit die septima (Heb 4:4, where the Clementine prints requievit Deus). ⭐ This is a case where the elided subject noted at §1.3 does real work: the antecedent I am relying on is a word Migne's lemma does not print, and the Clementine's does.
• 0649D qui mala retulerunt, non crediderunt — subject of crediderunt = the people, not the spies; the spies are the relative clause. Fixed by Num 13–14 and by Exploratoribus standing as the dative the belief is withheld from.
• 0650C quo deberet dari / nunquam post loqueretur — subject = David, carried from non loqueretur de alia David at 0650B.

⚠ One I could not settle and did not force: 0647D communicaverant carni … quae per sanguinem accipitur. The feminine quae can look back to carni (flesh, the member of the pair that "blood" stands for) or to anima in the intervening ex anima et corpore. Both construe. I rendered it flat — "which is taken by the blood" — supplying no antecedent, so the English is exactly as undecided as the Latin. Naming it rather than choosing.

§4 · THE SCOPE QUESTION AT 0650C — where nunquam attaches, and why I left it open

Trigger condition for test 2a fires here (a negative with a quantifier/correlative adjacent):

…requiem praestitisset nunquam, id est nullo tempore; posthac, id est in sequenti tempore; loqueretur de alia die, id est tempore gratiae, quo deberet dari. Vel, nunquam post loqueretur de alia, scilicet requie, praeter terram promissionis.

In the Clementine (Heb 4:8) numquam governs the apodosis — numquam de alia loqueretur, "he would never speak of another". Migne's serial gloss cuts it onto praestitisset and glosses it nullo tempore, which reads it with the protasis; then his own Vel clause puts it back on loqueretur. Migne prints both scopes and adjudicates neither.

So neither do I. I rendered the serial as a serial — "had given rest, never, that is, at no time; thereafter, … would he speak of another day, …" — with the comma before "never" holding it loose, and then rendered the Vel clause with the negative squarely on "speak". Supplying a scope in the first half would have been the 2a defect exactly: a negative correctly present, landing in the wrong place in English, passing every count. ⭐ Test 2a's own warning applies: it validates the mapping I propose, not the one the Latin makes. Here the sentence's structure (Vel) is what proves there are two mappings, so the honest English is the one that carries both.

§5 · THE FOURTEEN PUNCTUATION SITES THE BAND PASS RETURNED

Every one of these was a mark I had supplied or softened in draft; none is a Migne defect and none is reportable. Listed because test 3 exists to be run and its yield is the evidence it was:

Supplied by me, removed — a colon at 0643C personaliter, ut figura; a semicolon at 0644A omni creaturae, quae virtus; a colon at 0644C a nomine angelorum, hoc est nomen; a semicolon at 0645A in mundo erat, quae assumptio; a colon at 0646A funguntur officio, hoc est opus angelicum; two semicolons at 0646B in the fluimus / effluimus / pereffluimus chain (Migne sets it entirely with commas, and the chain is the argument); a semicolon at 0647A Claritate immortalitatis honore where Migne sets no mark at all; a semicolon at 0647B aggravat ipsum animum, sed hoc non in Christo; two semicolons at 0647D pueris, vel eisdem … carne et sanguine, et hoc similiter; a semicolon at 0648D Fabricavit dico, per se enim; a semicolon at 0651C per meliorem legem suam, et quomodo.

Deleted by me, restored — the colon of Dicit: loquens at 0643B; the colon of Quod vero ait: mercedis at 0646C; the colon of Apostolus dixit: apprehendit at 0648B.

⚠ 0646B and 0647D would each have cancelled against a total count — two supplied semicolons in one band, in a book that prints 31 semicolons of its own across my five chunks. Migne's own practice here is nowhere near zero, so Tobit's zero-semicolon exception is not available in this book and was not invoked (test 3's own caveat). No comma→semicolon sweep was run in either direction.

§6 · MIGNE'S OWN PRINTING — read as printed, logged, not tidied

6.1 · ⭐ 0649D — the repeated clause, and the one thing in my range I want a plate on

…ne quis putet satis esse audire de requie, quia [0649D] et illi omnes audierunt, sed non omnes audierunt, sed non omnes pervenerunt.

The sentence contradicts itself: omnes audierunt … sed non omnes audierunt. Heb 3:16's own sense (Quidam enim audientes exacerbaverunt : sed non universi) and the gloss's purpose — that hearing is not enough — both want the middle clause to say something other than audierunt. The likeliest shape is an eye-skip that has copied audierunt into a slot wanting a different verb (crediderunt, obedierunt), the second sed non omnes being the trigger.

Rendered exactly as printed — "they also all heard, but not all heard, but not all arrived" — because Pattern 7 has no exception and the self-contradiction is the evidence. No marker, and the reason matters:

• Not [d: …]: Pattern 11 wants the same run set twice. What repeats here is sed non omnes (correctly, in parallel clauses) plus the verb; the run is not doubled, so a [d:] would misdescribe the shape.
• Not [sic: …]: every word is a real word in a real slot, so there is nothing to carry untranslated, and the ship test forbids a plate claim I cannot stand behind. Our Latin is Corpus Corporum's transcription, not the plate, and dittography is precisely the class a transcription generates on its own.
• Not [cj: …]: I have two candidate verbs and no way to choose between them from the file. Pattern 18 needs a conjecture, not a shortlist.

⚑ Plate read requested: PL 114 col. 0649D. This is the highest-value single read in my range — it decides whether Migne printed a contradiction or our source manufactured one, and the answer changes which apparatus (if any) the site takes.

6.2 · Other printed oddities, read as printed with no marker

• 0644C et non pro illis, addit prae illis — pro illis against prae illis two words later, in a sentence whose whole business is prae. Rendered "and not for them". Declined a [cj:], and the rival is named: reading prae is attractive and probably right, but the clause is elliptical rather than assertive ("because he could have this, and not …"), so the faithful English does not assert anything false — it merely goes opaque, and Pattern 18's bar (18a: "the test is whether the English asserts something false, not whether it is odd") is not met. Elision is not a false claim.
• 0647A Qui autem portavit imaginem Christi dicuntur filii hominum — singular verb in the relative, plural in the main clause. English does have an exponent for this, so rendering it as "he that hath borne … these are rather called" keeps the disagreement visible and no marker is needed. ⚑ This is the boundary of Pattern 12's second test: the marker is the only honest treatment where English has no exponent (11632's signa … ardeat). Here it has one.
• 0648D Fabricavit dico against Heb 3:4's passive fabricatur. Real word, real slot, construable as "[someone] built it, I mean". Rendered as printed. Crux, no marker.
• 0648A Mors erat arma — singular subject, plural predicate. Rendered "Death was the arms", which carries it; arma is in any case a plurale tantum and this may be no defect at all.
• 0651D Est enim tempus quod non congruit. The work's last sentence in my range, and it simply stops, with no connection to the sentence before it — which is about there being no cause for despair now. Rendered "For there is a time which is not fitting", negation carried. Very likely the head of a thought the abridgment cut. Crux; nothing to mark.
• 0644B non Deum deformavit — deformare here in its literal sense, "to give a shape to", not "to disfigure". Rendered "did not give God a shape". Not a defect; recorded because the familiar English sense would invert the clause.

6.3 · Verse-number order

In sequence throughout my five chunks. Heb 1:1–14, 2:1–18, 3:1–16, 4:2–16, with gaps where Migne glosses nothing (Heb 2:3, 2:5, 2:11–13, 2:15, 3:6, 3:15, 4:1, 4:6, 4:10, 4:13). Gaps are not disorder; no reordering was needed and none was done.

§7 · REGISTER CALLS

• Thou / you (Pattern 17), decided by number, never by dignity. Singular thou at 0645C (Unxit te, O tu Deus Fili), 0645C (Et tu), 0645D (Manuum → "of thy hands"), 0646A (Sede → "Sit thou"; Pedum → "of thy feet"), 0647A (Visitas eum, Super opera manuum), and through the whole Augustinian address at 0647B (tibi videtur … non recte cogitas, nec attendis) — a singular reader, so thou, exactly as a singular brother would take it. Plural you at 0649A (modo estote molles), 0649B (corda vestra), 0649C (ut putetis, Debetis hortari) — the Hebrews addressed as a body. ⭐ The switch at 0649A is visible on the page and is the source's, not mine.
• [nt:] considered and not used. All 8 notes in my range are bare citations — (Joan. XIV), (Psal. VIII), (Matth. XXV), (Apoc. XXI), (I Cor. XV), (AUG.). Pattern 15 decides by content, and none is prose. They stay [n:], verbatim, unanglicized.
• Pattern 4 [f:] not applicable — runbook false positive 4: this is a lemma-and-gloss commentary, not a florilegium, however much inline scripture it carries. No tags.
• Column-band arithmetic — runbook false positive 1: the bands run 0643A→0651D with the normal irregularities and nothing breaks mid-word or mid-construction across an anchor. Not reported, and no [ed:].
• anthropopathos (0649B) kept as a plain transliterated technical term rather than italic Latin, because Migne prints it un-italicized and every italic span must be 1:1 with the twin. The register rule for untranslatable technical terms yields to the marker rule.
• Lemma expansion (convention 3), and the line I drew. Amplioris (0648C) is expanded to "Of ampler glory" because the bare stump is not English standing alone; Propter passionem (0647B), Differentius nomen, Ad quem angelorum, Manuum, Pedum, Nusquam, Eis are left as they stand because each is readable English as printed. ⚑ Per the brief's closing ruling I record the class rather than the instances: a clipped lemma is expanded only where the stump will not carry itself, and the defect the ruling names — conformation — is not in play in either direction here.
• Em-dashes: zero in all five chunks. Not sought and not avoided; comma, colon and full stop carried every construction. No faithful sentence was restructured to reach that.

§8 · SECTIONS READ CLEAN

Read against the Latin under test 5 ("does it ASSERT what he asserts?") with the counts already known clean, and returning nothing:

• 0000 entire (ARGUMENTUM, 0643A–B). The one thing worth recording is that verify-english.mjs warns its ratio is 1.61, above the 1.5 expectation. Confirmed by re-reading: the Latin is a chain of compressed ablative absolutes and gerundives (sciens … demonstrans, in hac epistola scribenda, cujus sensum et ordinem retinens) which English cannot carry without finite clauses. Unpacking, not padding.
• 0645A–0645C (the Angelos gloss on nature and office, and the Augustinian vocative/nominative note). Dense with negation — Nihil enim visibiliter fit, quod non … aut jubeatur, aut permittatur — and every negative was traced to its own host word under test 2: non attaches to the jubeatur/permittatur pair, not to fit.
• 0651A–0651C (the long Divisionem animae / Compagum / Medullarum exposition, ~450 words). No divergence, no defect, no marker. The passage's difficulty is philosophical, not textual.
• All ten non solum … sed / non … tantum / sicut … ita correlatives in my range, which is test 2a's named trigger population, each with the host proven from the sentence's own structure rather than from what read well: 0644C non solum Patre, sed etiam angelis · 0646C non enim solum omnis praevaricator · 0646D non simpliciter … sed per signa · 0646D non illis, sed Deo · 0647B sed hoc non in Christo · 0647B non carni suae solum, sed etiam menti · 0647C non pro potentia tantum … sed pro dilectione · 0647D non ei naturae … sed misericordiae · 0648B non angelicae, sed humanae naturae · 0651D non solum per hoc quod Deus … sed per hoc quod homo. All ten sit on the correct host; none floated onto its adjacent verb.
• deservit / deserit: zero occurrences in chunks 0000–0004. Grepped, not assumed.

§9 · FOR THE MERGE AGENT

1. This file covers 0000–0004 only. §1.1's table and §1.3's declines are the deliverable; fold them into the merged §1 without renumbering the columns.
2. 0651D is a shared band. My chunk 0004 ends inside it; the lemmata Omnis namque pontifex, Quoniam et ipse, Nec quisquam, Non seipsum clarificavit and the Mt 3 / Ps 2 citations at 0651D–0652B are not mine and are not walked here.
3. One plate read requested: col. 0649D (§6.1), plus a lower-priority pair at 0644B and 0646A for confessus (§2.1). Both [cj:] sites and all five [var:] sites should be gated against the plate before the work ships, per the standing rule.
4. Migne's Offensus/Infensus doublet at 0649B is worth a line on the work page: it is the one place in these four chapters where the recension visibly carries two readings of its own.

Cols 0652A–0660A (CAPUT V–IX). Agent stint: Latin chunks 0005, 0006, 0007, 0008. *This file is one of three; merge with the other agents' cruces-.md before ship.**

Markers fired in this stint: two [var:] (0656C, 0658A) and one [cj:] (0659A). Everything else below is a crux and asserts nothing about the plate.

1. Lemma walk — all 148 italic spans in lines 184–331 of data/briefs/8993-lemmata.txt

Every span that actually stands in chunks 0005–0008 was collated in place against sources/vulgate/clementine-flat.txt at the verse being glossed, not merely against the corpus. The 22 pre-scan ⚠ flags in my range are adjudicated first; the ✓ and single-word spans that I checked and found to agree exactly are recorded in §1c, because a findings list alone cannot be distinguished from a report by an agent that never looked.

1a. The 22 flagged spans, adjudicated

1b. Spans marked ✓ or "single word" that I checked in place and that do diverge

The ✓ mark is not clearance, and three of them are not clean at the verse being glossed:

• 0656C Cum esset, (VERS. 4) — Hbr 8:4 reads cum essent qui offerent secundum legem munera. Migne prints the singular Cum esset. Rendered "When he was," as printed. No marker: a lemma opening VERS. 4.-- here is being used as an address to the verse, the gloss immediately re-quotes 8:4's protasis (Si ergo esset) correctly, and singular/plural of the copula in a two-word clipped lemma is not a sense claim I can stand behind without the plate. Plate-read candidate** (col. 0656C).
• 0654D neque finem vitae habens. — agrees with Hbr 7:3 verbatim; the preceding Neque initium, is clipped from neque initium dierum, which is ordinary etc.-abridgment (Migne prints the etc. between them). Clean.
• 0655A In quo enim haec, / Et amplius, — agree with Hbr 7:13 / 7:15 verbatim. Clean.
• 0656B Capitulum. — Hbr 8:1 Capitulum autem super ea quae dicuntur. Clipped, not divergent. Rendered "The chief point," and the same English is held for Hic incipit capitulum two words later, so the gloss's identification of the lemma with the section survives in English.

1c. Spans checked in place and found to agree exactly (no divergence)

Chunk 0005: Omnis namque pontifex. · Quoniam et ipse. · Nec quisquam, · Hic est Filius meus dilectus, in quo mihi bene complacui (Mt 3:17) · Filius meus es tu, ego hodie genui te (Ps 2:7) · Tu es sacerdos. · Secundum ordinem Melchisedech, · Preces supplicationesque. · Exauditus est. · Didicit, · Omnibus obtemperantibus. · juxta ordinem Melchisedech, · Ininterpretabilis. · Etenim cum deberetis. · Omnis enim, · Verbum caro factum est, (Jo 1:14) · In principio erat Verbum? (Jo 1:1) · Quapropter. · Inchoationis Christi sermonem, · Fundamentum. · Impossibile. (×2) · Eos qui, · Renovari, · Rursum crucifigentes. · Saepe venientem. (×2) · Imbrem. · Generans herbam. · A quibus colitur. · Et fecerunt fructum nativitatis, (Ps 106:37, verbatim) · Confidimus. · Operis. · Cupimus. · Abrahae namque, · Nisi benedicens, · In quo abundantius volens. · Confugimus. · Quam sicut. · ac firmam, · Praecursor. · Exsultavit ut gigas ad currendam viam (Ps 18:6) · In via peccatorum non stetit (Ps 1:1).

Chunk 0006: Sacerdos Dei summi. · Neque initium, · In perpetuum. · Intuemini autem, · Et quidem de filiis Levi, · Contestatur. · Vivit. · Sub ipso. · Translato enim, · Translato enim sacerdotio. · Manifestum est enim, · Non secundum legem. · In tantum. · Melioris. · Hic autem, · Unde et salvare. · Sanctus, · Innocens, · Impollutus, · Segregatus a peccatoribus, · Qui non habet, · Hoc enim fecit semel seipsum offerendo. · Lex enim homines constituit,.

Chunk 0007: Talem habemus. · Sanctorum minister. · Quod, · et non homo: · Quod offerat. · Si ergo esset, · Responsum est, · Et melioris Testamenti, · Nam si illud prius. · Culpa vacasset, non utique. · Inquireretur, · Quod feci patribus eorum. · Non permanserunt in Testamento. · Hoc est Testamentum, · Dando leges: · Superscribam. · Non docebit. · Majorem.

Chunk 0008: Habuit. · Sanctum, · Saeculare. · Cherubim. · In anno. · Pro sua et populi ignorantia. · Verba delictorum meorum (Ps 21:2) · Nondum propalatam. · Sanctorum viam. · Quae parabola. / quae parabola est, · Usque ad tempus correctionis; · Per amplius. / per amplius, · perfectius, · In sancta. · Si enim sanguis. · Hircorum. · Ad emundationem carnis, · Immaculatum. · Viventi. · Intercedente. · Ubi enim. · Lana. · Et sine sanguinis effusione, · Ipsa autem coelestia. · Neque ut saepe, · Alioquin, · In consummatione. · Et quemadmodum. · Semel mori. · Post hoc autem. · Sine peccato apparebit. / sine peccato,.

2. The three markers, and why each rival lost

2a. [var:] at 0656C — fixit Deus for Vulg. fixit Dominus

Migne prints the divine name Deus at Hbr 8:2 where the Clementine prints Dominus, and he prints it twice: once in the closing-quotation span at 0656B (Deus et non homo.) and once as the glossed lemma at 0656C (fixit Deus,). The gloss then keeps Deus a third time in its own prose (quia et vetus posuit Deus, sed per ministrum hominem).

• Why it is marked and not merely cruxed: it is a substituted divine name in a proof-text, the class Pattern 14 exists for, and the three-fold internal agreement makes a one-off transcription slip in our source implausible.
• The rival, and why it lost: the rival was crux-only, on the ground that Deus/Dominus changes no argument in this gloss. It lost because the reader of our English is entitled to know that the words on Migne's page are not the words of the Vulgate at a verse the whole chapter is expounding; the sense-neutrality argument would suppress the divergence rather than record it.
• One marker, not three. It stands at 0656C, on the span Migne actually glosses.
• ⚠ Plate gate: this asserts what Migne's page reads. Col. 0656B–C needs a plate read before ship.

2b. [var:] at 0658A — Tabernaculum enim sanctum for Vulg. Tabernaculum enim factum est primum

The strongest lemma divergence in the stint. Hbr 9:2 opens Tabernaculum enim factum est primum; Migne's lemma is Tabernaculum enim sanctum. — three words replaced by one, and the substituted word (sanctum) is the very word the preceding verse's lemmata have just been glossing (Sanctum, … Saeculare. at 9:1, and the verse itself closes quae dicitur Sancta).

• Rival 1 (declined): read it as our source's loss and mark [ed:]. Rejected — [ed:] requires a demonstrated hole, and there is no hole here: the lemma is a complete, well-formed phrase and the gloss that follows attaches to it.
• Rival 2 (declined): crux only. Rejected because the divergence is material: a reader taking the lemma as the verse would carry away a "holy tabernacle" that Hbr 9:2 does not contain.
• Rendered as printed, "For the holy tabernacle."; the [var:] stands beside it and replaces nothing.
• ⚠ Plate gate: col. 0658A needs a plate read before ship. If the plate reads factum est primum, the marker is withdrawn and this becomes a Corpus Corporum patch.

2c. [cj:] at 0659A — Ac operibus mortuis for Ab operibus mortuis

Hbr 9:14 reads emundabit conscientiam nostram ab operibus mortuis, ad serviendum Deo viventi. Migne's lemma sequence for that verse runs Obtulit semetipsum → Immaculatum → Ac operibus mortuis → Viventi, i.e. it walks the verse word by word; the lemma stands exactly where ab stands.

• Ac is a real word, so Pattern 12 has nothing to wrap and addendum A forbids a [sic:] on a corpus argument. Pattern 18 is the right instrument: the printed word is real and the faithful English misleads, because ac makes the fragment a conjunction where ab governs the ablative that immediately follows it.
• The rival, and why it lost: the rival was crux-only, on the ground that the fragment reads tolerably either way and the gloss (id est peccatis) treats operibus mortuis as a substantive regardless. It lost because [cj:] is additive — Migne's Ac keeps its English ("And by dead works") and the conjecture stands beside it — so nothing is concealed and the reader is not left to reconstruct why a preposition became a conjunction. The a/b–a/c confusion is also the cheapest possible type error, which is what makes the conjecture worth printing rather than burying.

3. Cruces where I fired nothing

3a. ⛔ Two split-type candidates — the plate-read requests I most want honoured

Both are non-words as printed (addendum A's class, which has been right every time about an error existing) and both sit where a line break plausibly falls:

1. 0654A — quia patres suos, qui eos in fide genu erunt, imitantur. Read genuerunt, "begot". As printed, genu erunt is "knee they-will-be", which is not Latin here.
2. 0655D — Sicut Christus per manet, ita et lex quam attulit. Read permanet, "remains".

Why no [sic:] on either, and this is a deliberate departure from Pattern 7's non-word ruling. Hard rule 8a (2026-08-28) is later and governs: a [sic:] accuses Migne's type, and our Latin is Corpus Corporum's transcription, not the plate. A space inserted at a line break is the single most likely digitization artifact there is, and the 2026-08-23 batch withdrew seven false [sic:] that each publicly blamed Migne for exactly this class. The ship test therefore applies: unsure, cannot settle it from the file, so the crux is the safe instrument. Both are rendered for their evident sense ("who begot them in the faith", "As Christ remains"), which is what test 1 requires — nothing is swallowed. ⚑ If a plate read confirms the split on the page, both become [sic:]; if not, both become data/tei-patches/8993.json entries. Cols 0654A and 0655D.

3b. 0652B — propter carnem absumptam

Christus … non secundum id quod natus est de Patre … sed propter carnem absumptam, propter victimam, quam obtulit pro nobis. The doctrinal commonplace is carnem assumptam, "the flesh he assumed", and the whole sentence is about the incarnation.

Declined, and here is why the conjecture lost the marker rather than the argument. absumptam (absumo, "consume, use up") is a real, well-formed word, and it construes in place: the parallel member is propter victimam quam obtulit, and a victim's flesh being consumed is not nonsense in a sacrificial sentence. Addendum A is 10 for 10 that a real word which reads oddly is Migne's until a plate says otherwise, and both readers who raised refuted real-word cases in that batch were commended for asking for the plate instead of patching on the argument. Rendered as printed, "on account of the flesh consumed". Plate-read candidate, col. 0652B. ⚠ Note also brief §4/addendum C: if the plate reads assumptam, the interesting finding is not the typo but that our file has the harder word — the normalization class runs in both directions.

3c. 0655A — per hoc quod accipiunt sacrificium

Hbr 7:5, the verse being glossed, reads de filiis Levi sacerdotium accipientes, mandatum habent decimas sumere. The gloss paraphrases it with sacrificium.

Declined. I raised it as a [cj:] candidate and then withdrew it: the printed reading is not merely a real word, it is true of the Levites — priests do receive the sacrifices (Num. 18), and "because they receive the sacrifice … they have a commandment in the law" is an argument, not a wreck. Pattern 18's bar is that the English asserts something false; this English does not. Rendered as printed.

3d. Migne's punctuation, preserved where it hurts

Three places where the printed mark makes the English read as an error, all carried through per Pattern 8 with the mark in its printed position (band parity is exact for all four chunks):

• 0653D–0654A — Vel illi a quibus colitur. sunt illi iidem qui se colunt in bonis operibus. A full stop stands between the relative clause and its verb. Carried: "Or those by whom it is tilled. are those same men who till themselves in good works". The lower-case "are" is deliberate — it mirrors Migne's lower-case sunt; supplying a capital would be tidier and would hide the mark's position.
• 0655A — Immoratur in commendanda dignitate. Melchisedech et excellentia sacerdotii ejus; The same defect: dignitate Melchisedech is plainly one phrase and a full stop splits it. Carried: "He dwells upon commending the dignity. Melchisedech and the excellence of his priesthood;".
• 0658A — the VERS. 2 gloss ends with no terminal stop: intus significat Novum. Carried; the English paragraph ends "within signifies the New" with no full stop. Never supply.

Also carried without comment in the text: Distat inter sacerdotium, et sacerdotium (0655C) keeps its comma; sicut et Noe, Abel Abraham (0654D) keeps its missing comma between the two names.

One comma NOT carried, and it is the only mark in the stint I could not place. 0659D–0660A prints quod aliqua peccati sui, necessitate Christus oblatus sit. aliqua … necessitate is a single ablative phrase and English word order cannot admit a comma inside it without producing gibberish. Rendered "that Christ was offered by any necessity of his own sin,". Logged here rather than adjudicated in the text; comma parity is not a checked class, so nothing is concealed by this being a crux.

3e. 0653B — Non rursus, id est non olim

The gloss on Hbr 6:1 reads Non rursus, id est non olim, ut sicut prius ceciderunt, iterum cadant. "Not again, that is, not formerly, that as they fell before, they should fall again" — the ut clause states the outcome the non is refusing, so the sentence reads as a self-contradiction. olim is a real word and construes ("not as of old"); no repair is attempted and none is licensed. Rendered as printed, with both negations carried. Candidate reading, not marked: non olim for non iterum / non ita*. Plate-read candidate, col. 0653B.

3f. 0656C / 0656D — a sentence printed twice, with its nouns swapped

Jam a sacerdotio et sacrificio transit ad differentiam Testamentorum closes the VERS. 4 gloss; Jam a sacrificio et sacerdotio transit ad differentiam Testamentorum opens the VERS. 6 gloss, two paragraphs later. Not Pattern 11 dittography — the runs are not adjacent, they are not identical, and each sits in a different verse's gloss, so this looks like the recension repeating itself rather than a compositor doubling a line. Both rendered in full, in place, with the noun order each one prints.

3g. Unexpressed subjects and objects I had to supply

Per the brief, each is named with what decides it:

• 0653A Nec tamen sic deserant — "Nor yet let them so forsake them." Subject = spirituales from the preceding clause. The object is unexpressed; I read it as the carnal (carnalibus, the dative of the immediately preceding ista carnalibus non omnino taceant) rather than ista, because the following ut volentes ea perducere already carries ista as ea, so the two would collide. ⚠ Both construe. If a later reader prefers "forsake these things", the sentence is unharmed.
• 0655C quod quia vetus lex non potuit, exclusa est — exclusa is feminine singular and agrees with lex, so the thing shut out is the old law, not the grace. Rendered "which, because the old law could not give, it was shut out."
• 0655D sicut in lege fuerunt multi, quia mortales — "because they were mortal"; the copula is unexpressed and the subject is multi (the many priests), not the law.
• 0658B Christus pro populo suo obtulit, pro se vero non, nisi in membris suis — the elided verb is obtulit. Rendered "but for himself he did not, save in his members", keeping non on the verb.

3h. Negation attachment — the sites that fired test 2a's trigger condition

Every printed negative in the four chunks was walked to its host. Four sites had a quantifier, correlative or pivot adjacent to a plausible verb and were proved from the sentence's own structure:

• 0653A ista carnalibus non omnino taceant — host is omnino taceant as a unit ("do not altogether keep silence"), not taceant alone, which would say they keep silence entirely. The propter catholicam fidem, quae omnibus praedicanda est that follows requires that they do speak.
• 0654D Christus omnino non habuit initium neque finem — host is habuit; omnino intensifies the negation rather than being negated. Rendered "absolutely did not have a beginning nor an end", with neque carried separately. (This is the 8960 @0734B class; English word order was checked for scope reversal.)
• 0657C justi non extra terrentur, sed intus diligunt justitiam — the sed correlative fixes the host on extra, not on terrentur. Rendered "are not terrified from without, but within they love justice." (The 8960 @0740B model, handled correctly.)
• 0659B in cujus morte constat terrena non promitti, sed aeterna — the sed aeterna pivot fixes the host on terrena. Rendered "it is certain that not earthly things are promised, but eternal", not "earthly things are not promised", which would be the same words in the wrong scope.
• 0659A non modo gentilium, sed et earum praevaricationum — ordinary non modo … sed et; the second non four words later (quas lex ostendebat, non solvebat) attaches to solvebat and stays there.
• 0654A velint semper eadem ipsa discere — ⚑ checked against the 8987 @0624B defect explicitly. There is no negation in or near this clause; velint is rendered positively, "they wish always to learn these same things."

3i. deservit / deserit — checked, clean

The standing confusion pair does not occur in this stint. The only deser- form in cols 0652A–0660A is 0653A deserant (desero, "forsake"), which takes an accusative and is rendered as such; deservio does not appear. (The work's deserviunt / deseramus are in chunk 0012, outside my range.)

3j. usque ad — checked per occurrence (addendum D)

Two occurrences in my chunks, and both are the abridgment formula, not the verse's own words: 0654A ([n: (AUG.)] Jacta in Deum curam tuam, etc., usque ad si ad tempus fluctuat…) and 0656B (semper Magistrum suum sequens, etc., usque ad cui oportet fungi munere sacerdotis). Both are rendered "as far as" per the corpus convention, and no verb is supplied to complete either resumption tail; each tail was parsed as ordinary prose. ⚠ A third, 0658A Usque ad tempus correctionis;, is a lemma — Hbr 9:10's own words (usque ad tempus correctionis impositis) — and is rendered as such, "Until the time of correction;". Letting the formula reading swallow it is precisely addendum D's trap.

3k. Sections read clean

Read with no candidate raised: 0652C (VERS. 7–8), 0653C (the six-fold foundation), 0654B–C (VERS. 10–17), 0656A (VERS. 25–27), 0657A (VERS. 7), 0657D (the Majorem gloss), 0658B–D (VERS. 7–12), 0659C (VERS. 22–23), 0659D–0660A (VERS. 26–28, apart from the comma at 3d). The heads (CAPUT V. … CAPUT IX.) take the corpus's CHAPTER V. … CHAPTER IX. form; no head text diverges.

4. What I want a plate read on, in priority order

1. 0654A genu erunt and 0655D per manet — split type (§3a). These decide two markers.
2. 0658A Tabernaculum enim sanctum — gates the [var:] at §2b.
3. 0656B–C Deus for Dominus (twice) — gates the [var:] at §2a.
4. 0659A Ac operibus mortuis — supports the [cj:] at §2c.
5. 0652B carnem absumptam — §3b; a normalization check as much as a defect check, so read the whole line and its neighbours (addendum C).
6. 0653B id est non olim — §3e.
7. 0656C Cum esset for cum essent — §1b.

⚠ Our PL 114 copy clips a margin on some pages in this stretch. Nothing in this stint was recorded as illegible; if any of the above proves unreadable, data/volumes.json lists a second witness for the tome and the check costs one request.

Cols 0660B–0670D (Heb 10–13, the tail of the epistle with the closing salutations). Agent range: Latin chunks 0009/0010/0011/0012 → English of the same names. This file is one of three; the merge into cruces.md is a required later step.

Comparand throughout: sources/vulgate/clementine-flat.txt (Clementine), cited by the source's own abbreviations (Hbr, Ex, Mt, Gn, Jo).

1. Markers fired — the complete list

No [sic: …] was fired anywhere in this batch. Four non-word candidates were found and all four are logged in §3 as digitization-corruption candidates rather than plate defects. The reason is stated once, in §3, and it is a deliberate application of the ship test.

No [nt:], no [ed:]. All five [n: …] notes in the batch (CHRYS., AUG. ×3, and the scripture locators Joan VII · Exod. II · Matth. X · Matth. V) are citations or patristic extracts introduced by a siglum, not editorial prose; they pass through verbatim under convention 4 and Pattern 15.

2. The lemma walk — 7a″. Every italic span in cols 0660B–0670D

data/briefs/8993-lemmata.txt lines 331–510 fall in this range. All 183 spans were read against the Clementine at the verse actually being glossed. The eighteen ⚠ flags first, then the agreements, then the single words.

2a. The eighteen pre-scan ⚠ flags — every one checked in place

2b. The usque ad audit (addendum D) — four occurrences, decided one by one

• 0663A [n: (CHRYS.)] … etc., usque ad ut autem nos incidatis… — formula, italic, between an etc. and a resumption tail. Rendered as far as (8985 precedent).
• 0663B Vivit. [n: (AUG.)] … etc., usque ad et deficientem sublevat — formula.
• 0664A [n: (AUG.)] … nec tamen, etc., usque ad priusquam… — formula.
• 0670A Confidentium. [n: (AUG.)] … etc., usque ad cujus invisibilis sacrificii… — formula.
• 0666A a sanguine Abel usque ad sanguinem Zachariae — NOT the formula: roman, not italic, and the words are the allusion's own (Mt 23:35). Rendered "from the blood of Abel unto the blood of Zacharias".
• 0666B non acceperunt statim promissam vitam usque ad Christum — NOT the formula, ordinary preposition. Rendered "until Christ".
• No verb was supplied to complete any resumption tail.

2c. Spans checked and found to agree with the Clementine exactly

Read in place, verse by verse, and rendered without conformation. Hbr 10: Umbram enim · Non ipsam imaginem · Accedentes · Cultores semel, · Commemoratio, · Noluisti · Corpus autem aptasti mihi · Non placuerunt (⚠ tibi dropped — abridgment) · Tunc · Dixi: Ecce venio, · Superius dicens · Tunc dixi · Sanctificati, · Et omnis, · Sedet in dextera · Scabellum · Contestatur autem nos, · Postquam · dando leges, (⚠ meas dropped — the Douay's "my laws" was deliberately NOT written) · et peccatorum et iniquitatum eorum jam non recordabor amplius, · Superscribam · scribam, · superscribam, · Ubi autem, · Per velamen · In plenitudine fidei · Fidelis enim · Et consideremus, · Sicut consuetudinis · Sed consolantes · Voluntarie enim peccantibus · Post acceptam notitiam veritatis, · voluntarie, · Jam non relinquitur pro peccatis hostia; · Duobus vel tribus testibus · Conculcaverit, · Pollutum duxerit · Et spiritui gratiae · Mihi vindicta, · Rememoramini · rememoramini, · Pristinos, · Magnum certamen, · Et in altero quidem, · Spectaculum · Vivit. · Nos autem,. Hbr 11: Est autem fides · substantia, · Argumentum, · Senes: · Fide intelligimus · Fide plurimam · Defunctus adhuc loquitur · Quia est · Fide Noe, · Et justitiae, · Obedivit, · Quoniam fidelem · Ab uno · Et sicut arena · Et salutantes (×2) · Hospites · Et si quidem, · Cum tentaretur · Et in parabolam · Isaac · Adoravit fastigium, · De ossibus suis mandavit · Improperium Christi · Daemonium habes (Jo 7:20 / 8:48, both attest Daemonium habes) · Non veritus · Fide celebravit Pascha · Transierunt mare Rubrum, · Circuitu dierum septem · Excipiens exploratores cum pace, · De Gedeon · Samson · Jephte · Qui per fidem · Repromissiones / repromissiones · Alii autem distenti, · Quibus dignus non erat mundus · Deo pro nobis melius aliquid providente. Hbr 12: Ideoque et nos · Nubem · Qui proposito sibi gaudio, · Deficientes · Omnem filium · Ergo adulteri · Reverebamur · Et illi quidem · Ad id quod utile est, · in recipiendo sanctificationem ejus · Disciplina · Manus · genua, · Gressus · Sanctimoniam · Beati mundo corde, quoniam · Sine qua nemo videbit Deum · sine qua nemo, · Contemplantes, · Ne qua radix, · Radix amaritudinis, · Ne quis fornicator, · Primitiva · Non enim accessistis (×2) · Turbinem · Caliginem · Procellam, · Tubae sonum · Vocem verborum · Non enim portabant · Et si bestia, · Sed accessistis · Ad Sion · Et civitatem Dei · Qui conscripti sunt (×2) · Adhuc semel, · semel · Quae sunt immobilia · Itaque regnum immobile · Deus noster ignis. Hbr 13: Charitas · Sint mores · Sine avaritia · Non te deseram · ut confidenter dicamus, · Neque derelinquam · Non timebo, · Mementote, · Jesus Christus, · Non te deseram, neque derelinquam · Habemus altare, · de quo edere non habent, (⚠ potestatem dropped — abridgment) · Tabernaculo, · Quorum enim, · Animalium · Portam · Per ipsum ergo · Beneficentiae (×2) · Non deserentes collectionem · Ut cum gaudio · Non expedit vobis · Orate pro nobis · Bonam conscientiam habemus, · In omnibus · Faciens, in vobis, · Etenim perpaucis,.

2d. One non-italic scripture quotation, checked because guillemets carry the same risk

0664A « Exi de terra tua, et de cognatione tua. » vs Gn 12:1 Egredere de terra tua. Exi for Egredere, sense identical; the rest verbatim, truncated before et de domo patris tui. No marker — Pattern 14 needs a change of sense. Rendered "Go out from thy land". The single guillemet pair is reproduced 1:1.

3. Non-word candidates — four, and why NONE takes a [sic: …]

Addendum A measured the split 10 for 10: every suspicion aimed at a non-word was confirmed as our corruption, i.e. Corpus Corporum's, with Migne's plate clean. A [sic: …] is a public claim that Migne's plate is defective. Firing four of them here would therefore mean publishing four claims I expect a plate read to withdraw — precisely the seven-withdrawn-[sic:] failure of the 2026-08-23 batch. Under the ship test the crux is the safe instrument and it is what I have used.

What is in the English instead: each printed form is carried into the English untranslated, in italics, in its own grammatical slot, with no conjectural sense supplied around it. Nothing is patched and nothing is silently repaired. ⚠ I record openly that this leaves the reader an unmarked italic Latin word, which Pattern 12 says normally signals an editorial choice rather than a defect. If the plate read comes back showing Migne prints these forms, all four should take a [sic: …] in one commit. That is the correct resolution and it is cheap; the reverse (withdrawing four false markers) is not.

3a. Suspiciously tidy forms (brief §4 / addendum C) — read the whole line

Two lines were examined closely for a neighbouring word while checking the target, per addendum C, and both came back clean: 0663A's whole CHRYS. sentence around securint, and 0667B's whole Gressus paragraph around sanctur. Nothing in either reads tidier than its surroundings. ⭐ Recorded because a checked negative is worth having.

4. [cj: …] — 0661D vetata

VERS. 20.-- Per velamen. Fides nostra est vetata, quia fides non habet meritum, cui humana ratio praebet experimentum.

vetata is a perfectly good participle of veto — "forbidden" — so Pattern 12 has nothing to wrap and Pattern 7 renders it. And the rendered page then says "Our faith is forbidden", which is not what the gloss asserts. This is Pattern 18's founding shape exactly (munde / mundo at 11081 @0030B).

• The conjecture is velata, "veiled". The lemma being glossed is Per velamen, and the clause that follows is Gregory's commonplace that faith has no merit where human reason supplies proof — i.e. faith is veiled, and is meritorious precisely because it is.
• The rival, and why it lost. vetata read as "forbidden" was tested against a possible sense "our faith is fenced off / made inviolable" — it fails, because veto takes a person or an action as its object, not a thing held back from view, and because that reading leaves Per velamen doing no work in its own gloss.
• The gloss is additive (Pattern 18): Migne's word keeps its English and the conjecture stands beside it. Corpus frequency (vetata 1 · velata 0 · velamen/velamine 51) was not used to fire this — addendum A rules frequency worthless for real words. The argument is the sentence's own.

5. Dittography — 0661A

Ideo venio [0661A] ad implendam tuam voluntatem tuam…

Two identical copies of tuam, not adjacent, straddling voluntatem. Per 11a the marker sits on the INTRUSION, not on the second copy, and the intrusion is the first: Heb 10:7 reads voluntatem tuam and the gloss quotes it in that order eight words earlier (ut faciam et compleam voluntatem tuam), so the postposed copy is the true text and the preposed one is a false start with no slot of its own. English: to fulfil [d: thy] thy will. Looks like a compositor's doubling, not a textual one.

6. Real words that read oddly — rendered as printed, no marker (addendum A)

These are Migne's until a plate says otherwise. Each is logged so a blind reader knows it was seen.

• 0660B sed quare? vel semel? — two question marks on what is barely a sentence. Both preserved (Pattern 8); the batch's ? parity is 3/3 · 0/0 · 1/1 · 1/1.
• 0660D Celebrant figuras, figurae rem multi scientes, sed plures ignorantes quae cessare debent re exhibita. Construes only loosely; rendered clause for clause without smoothing.
• 0661A in consilio Deitatis, quae est caput mei — genitive pronoun where a possessive is expected. Rendered "which is my head". Pattern 9 (no English exponent) borders this.
• 0661B non pervenit ad consummationem, sic etiam omnis sacerdos — Migne's comma where English wants a stop. Comma kept, per Pattern 8 and per the standing warning not to sweep comma→semicolon.
• 0661C quod non valet ad probationem nostram — the negation is on valet and it is carried, even though it makes the Apostle cite what does not prove his point. No repair.
• 0662A nisus seu nisio — nisio is a rare nominal from nitor; a real formation. Rendered "the effort, or the striving".
• 0662B Quod si non facit judicium, te exspectat. ⚠ Two readings, and the marker took neither because none was fired. Migne's comma falls after judicium, which yields "if he does not do judgment, it awaits thee" — but exspectat then has no subject at all. Read with the comma one word earlier (Quod si non facit, judicium te exspectat) the sentence is ordinary: "if he does not do this, judgment awaits thee." The English takes the second, because the first leaves a subjectless verb, which is a grammatical failure and not merely an awkwardness. The comma is a comma, not a sentence-type mark, so Pattern 8 is not engaged. Logged rather than marked.
• 0665A et caeteri et patres — "and the rest and the fathers", a doubled connective in a list. Rendered as printed.
• 0665B Noluit ibi manere — ⚠ ibi is ambiguous between Egypt (the danger he left) and Madian (where he in fact stayed). Egypt fits the argument, since remaining in open danger is what would tempt God. No supply was made: English "there" is exactly as ambiguous as Latin ibi, so the reader inherits the choice rather than ours.
• 0666B Cur non subditur: — rendered "Why not, is subjoined:", keeping both words and the colon. Migne prints no question mark and none was supplied.
• 0666B qui nos tandiu exspectaverunt requiem — two accusatives after exspectaverunt. ⚠ Two readings. The English took "who so long awaited us and rest", reading both as objects. The rival — "who so long awaited rest for us" — reads better and fits the argument (the fathers waited; we receive at death), and it lost because it requires nobis, i.e. it is an emendation dressed as a translation, which Pattern 7a forbids.
• 0670C qui homo ab officio, eo scilicet quod hominem decipiat — the devil called a man "from his office". Rendered as printed.

7. Negation and attachment — the tests, run

Test 2 / 2a. Every printed negative was traced to the single Latin word it attaches to. The trigger population (a negative on a verb with solus / omnis / omnino / solum / tantum or a non solum … sed / sicut … ita pivot adjacent) fired nine times and all nine were proved from the sentence's own structure, not from what reads well:

Test 1 (a [sic:] swallowing its word). Vacuous here — no [sic:] was fired. The equivalent check was run on the four carried non-words instead, asking 7a⁗'s three questions and the fourth one the brief adds (does the clause still have its OBJECT?): securint takes the clause's only verb and the hole is left visible with its object alios intact; padebat is the verb of an impersonal whose accusative eum and infinitive ferre both survive in English; sanctur is a bare subjunctive with no object; serviendium is a gerundive whose dative ei survives. Strip every marker from the four English chunks and no stranded article or preposition remains.

Test 2b (an anchor present, in order, in the wrong place). All 42 anchors were checked against the WORD they stand beside in the Latin, not merely for presence. Three sit mid-construction and were split to match Migne rather than moved to a sentence boundary: a terra [0664B] sua → "from his [0664B] own land"; gravia et indigna [0667A] perpetitur → "grievous and unworthy [0667A] things"; timore tribulationis [0666D] fidem negare → "through fear of tribulation, [0666D] to deny the faith". The lemma-internal [0669B] is treated in §2a.

Test 3 (punctuation by band, never by total). ? : ; were compared band by band across all four chunks; every mismatch was traced and repaired, and the only residual differences are colons and semicolons inside [d:], [cj:] and [var:] markers, which are ours by construction. Eight supplied semicolons and two supplied colons were found this way and reverted to Migne's commas — every one of them would have passed a total-based check because the totals were not equal anyway.

Test 5 (does it ASSERT what he asserts?). Run last, paragraph by paragraph, with the counts already clean. It is what produced §4 (vetata), §6's 0662B comma reading, and §7's 0666A omnino scope. ⚠ The smooth-where-the-Latin-is-rough suspects were 0660D's Celebrant figuras and 0666B's nos … requiem; both are logged unrepaired rather than tidied.

8. The unexpressed subject — every supplied pronoun, named

• 0660C quod non vetus → "which the old victim does not". vetus is the neuter adjective standing for hostia vetus, from Una hostia Christi in the same sentence.
• 0660C Nunquam ea ante destruxit → "He never destroyed them before". Christ, from Probavit quod Christus fuerit semel mortuus at the head of the chapter and from in templo cum hostiis praesentatus in the same clause.
• 0662C et ita pollutus est et nos polluit, si… nos retrahit → "he is polluted and pollutes us". ⚠ Two candidates: sanguis Christi (the nearest nominative) and the man of Talis est qui…. The man wins from the structure: Christ's blood draws us from carnal observances, so it cannot be the subject of nos retrahit away from unde est remissio. Naming the blood would have made the page say Christ's blood pollutes us.
• 0664A Prius risit → "She laughed at first". Sara, from Hbr 11:11's Fide et ipsa Sara, not from the sentence; the gloss never names her.
• 0665D Ipse, accepto signo velleris → "He", Gedeon, from the lemma De Gedeon immediately above. Likewise Hic → Samson, Hic → Jephte, each from its own lemma.
• 0669B ut Loth nesciens esse angelos, sciens tamen quis in eis esset → "as Lot, not knowing that they were angels, yet knowing who was in them". Both plurals are the quidam of the lemma.

Pattern 17 (thou = singular). te terret / te exspectat / exceptus es → thou (singular in the Latin). vobis · vestra · estis · erigite · dubitetis · Salutate → you / ye (plural). The switch at 0667C is Migne's own — volens te amatorem boni esse inside a paragraph otherwise addressed to the plural — and it is reproduced, not levelled.

9. deservit / deserit — the standing check, discharged

Five instances in range, all correct as printed; no confusion found in either direction.

• 0669D qui tabernaculo deserviunt — deservio + dative tabernaculo. ✓
• 0670A tabernaculi delicias deseramus — desero + accusative delicias. ✓
• 0669A Non te deseram and 0669C Non te deseram, neque derelinquam — desero + accusative te. ✓ (Hbr 13:5, Clementine verbatim.)
• 0670B Non deserentes collectionem — accusative. ✓ (Hbr 13:16… see below.)

⚠ One related observation, offered as a crux and not as a finding: at 0670B the gloss cites Hoc supra dixerat: Non deserentes collectionem, which is Hbr 10:25, not the Hbr 13:16 verse under discussion — and supra is exactly right, since 10:25 was glossed in this same batch at 0661D. The reference is sound; noted only because a collectionem / deserentes pair is the shape a checker will stop on.

10. Migne's own printing, not tidied

• 0663A VERS. 31.-- Rememoramini. — the address is VERS. 31 but Rememoramini opens Hbr 10:32; 10:31 is Horrendum est incidere in manus Dei viventis, whose words appear in the CHRYS. resumption tail immediately below (ut autem nos incidatis in manus Dei). Migne's numbering. Reproduced as printed, not corrected.
• 0663A the paragraph Invitat eos per antefacta ends with no terminal stop. None was supplied (Pattern 8).
• 0665A [n: (Exod. II)] Quod tamen patienter sustinuit — no full stop after the note. Reproduced.
• 0669A/B the chapter head is VERS. 1, 2.--, a doubled verse address. Reproduced.
• Chapter heads render CAPUT X./XI./XII./XIII. → CHAPTER X./XI./XII./XIII., matching the Roman-numeral form this work's own shipped chunk 0005 uses (CAPUT V. → CHAPTER V.), not Ruth's spelled-out CHAPTER ONE. A rendering found in another book is a hint, not an authority; the work's own precedent governs.

11. Plate reads wanted (in priority order)

1. 0667B sanctur — with 11065 @0229D sanctur in the same look. If Migne prints sanetur at both, this is a named systematic e→c transcription fault worth a corpus sweep, not two isolated repairs.
2. 0664A Fide qui vivit (Hbr 11:8) — does the plate read vivit or vocatur Abraham? The [var:] stands either way; a vocatur plate would additionally justify withdrawing it in favour of a source patch.
3. 0663A securint · 0665A padebat · 0669D serviendium — if any prints as given, that instance takes a [sic: …].
4. 0661D vetata — if the plate reads velata, the [cj:] is withdrawn and the line becomes a source patch instead.
5. 0670A Confidentium (Hbr 13:15) — a real word, so no [sic:] is proposed; but confitentium → Confidentium is a plausible OCR path and the [var:] would then be ours to withdraw.

⚠ Nothing in this range was illegible in our copy, so no margin-clip is reported.

12. Sections read clean

Read against the Latin with no crux raised and none needed: 0660C–0660D (VERS. 2–6, the medicine simile and the figures) · 0661B–0661C (VERS. 10–16) · 0662C–0662D (VERS. 29–30, the Chrysostom consolation) · 0664B–0664D (VERS. 11–19, Sara, the stars and the sand, the sailors' salutation) · 0665B–0665D (VERS. 22–32) · 0666C–0667A (VERS. 1–6, the scourging of the only-begotten) · 0667C–0668C (VERS. 14–23, the Sinai/Sion contrast, the densest exhortation in the range) · 0669C–0670A (VERS. 6–12) · 0670B–0670D (VERS. 16–22, the closing salutations).

Candidates raised and rejected, with reasons, so they are not re-raised: Tanquam (orthography, not a variant) · coelum/coelos (orthography) · caetero (orthography) · aquiescas-class medieval spellings generally (Pattern 9) · propositos at 0670B (attested by-form — §2a) · securint as an attested syncope (§3) · vetata on a corpus-frequency argument (§4 — the marker rests on the sentence, not the count) · the usque ad at 0666A and 0666B as the abridgment formula (§2b) · every Pattern 4 [f:] inline-citation tail (runbook false positive 4: this is a lemma-and-gloss commentary) · every column-band gap (runbook false positive 1).