Bipartite: does the vault's own independent_roots mechanism confirm or undercut the claim that its source-tier apparatus revives the système de preuve légale run in reverse?

legal-historyvault-epistemologyindependent-rootspreuve-legalesource-disciplinecross-domain-bridgemeta-analysiscorroboration

Neither target note cites the other, and a direct search of 10-inbox/raw/ for a prior capture naming both slugs together found none — observation-vault-source-tiers-revive-preuve-legale-run-in-reverse appears only in its own 2026-07-12 promotion lineage, and claim-independent-roots-is-bookkeeping-not-a-categorical-bar appears only in its own 2026-10-19 lineage. The closest prior art is one step removed: the 2026-10-19 capture that produced the independent_roots note compared that field against moc-one-witness-is-not-enough (the unus testis, nullus testis doctrine cluster) and concluded the resemblance is "real but not identical... a shared diagnostic concern, wired to two structurally different enforcement mechanisms." That prior work never touches the "run in reverse" observation note at all — a different, narrower pairing than the one this topic names. This capture triangulates the two notes the topic actually points at, using that prior verdict as a tool rather than re-deriving it.

The less-familiar side, read cold: independent_roots is a count recorded on a claim-note, hard-gated at exactly two checkpoints (promotion from review to candidate; the single-source concentration cap), otherwise just an honest label. The more-familiar side: the whole source-tier apparatus is read as a structural revival of the medieval système de preuve légale, with the vault's practice of refusing a same-mirror re-read as corroboration cast as the modern echo of Roman-canon law's two-witness requirement.

Claim: the specific mechanism Note A cites as evidence of a revived "independence rule" did not exist in named, spec-level form when Note A was written — it was formalized roughly eleven weeks later

Claim type: historical (dated facts about the vault's own documents, recoverable by direct read). Floor: Tier 3-4 acceptable for uncontested historical claims — achieved Tier 1 (both documents' own frontmatter and the operating spec's own changelog annotation, read directly).

observation-vault-source-tiers-revive-preuve-legale-run-in-reverse carries date_created: 2026-07-12 and grounds its "independence rule" claim in practice that existed at that time: the capture-verified / verified-verbatim ladder, and the observation that "a second reader of one carrier is 'one witness in disguise.'" The operating spec's independent_roots field — the mechanism claim-independent-roots-is-bookkeeping-not-a-categorical-bar examines directly — carries its own dated annotation: "(Stochastic-design proposal §3b, applied 2026-09-28.)" Eleven weeks separate the two: Note A's isomorphism claim was drafted against an informal practice, and the named field it is now most naturally read against did not yet exist. The field's own text states what prompted it: "This is the W39 lesson (Fel'dbaum's professorship, the Liang Zhongtang cluster: agreement between secondaries is a fact about their shared ancestor, not about the world) made into a field a lint can count" — a lesson drawn from two specific citogenesis cases, not from the legal-history comparison Note A draws. The two notes are not describing the same artifact at the same moment; Note B examines the settled, later, formal version of something Note A had only informally sensed.

Claim: read directly, the vault's actually-implemented corroboration mechanism is narrower and softer than the flat, general two-witness bar Note A's language evokes

Claim type: technical-mechanism (how the vault's own spec works, not merely that it exists). Floor: Tier 1-2 required — achieved Tier 1 (the operating spec, read directly; already established as claim-independent-roots-is-bookkeeping-not-a-categorical-bar, linked rather than re-derived here).

Note A states the parallel in categorical terms: "Roman-canon proof required two witnesses to the same fact, and the vault explicitly refuses a cross-model re-read of the same mirror as corroboration." Read against the spec's own text, the vault's actual corroboration bookkeeping — "count the distinct primaries those sources reduce to and record it as independent_roots: <n>" — is not a general refusal. It gates only at promotion (review → candidate) and at the single-source concentration cap; "outside those two checkpoints, a single-root claim is simply a single-root claim, flagged and left standing." A note can carry independent_roots: 1 indefinitely at status: seedling without being refused anything. This is a narrower claim than "the vault explicitly refuses... as corroboration" suggests on a plain reading: the refusal is local to two chokepoints, not a standing categorical bar comparable to a court that will not convict on one witness, full stop.

Claim: the historical système de preuve légale itself was not the flat two-witnesses-always rule the comparison implicitly reaches for — it graded evidence into half- and full-proof and allowed two half-proofs to sum, except at named, narrow chokepoints of its own

Claim type: historical (how the medieval system actually worked, already vault-sourced). Floor: Tier 1-2 required for the core mechanism — achieved Tier 2 for the half-proof/full-proof grading itself (Bloemberg, AJLH, already verified verbatim in the vault); the specific "except in grave criminal cases" exception clause remains at the vault's own existing flag, [unverified -- needs primary], carried forward unchanged here, not strengthened.

The medieval regime Note A compares the vault to was itself a graduated accounting system, not a bare "two witnesses or nothing" rule. A single eyewitness or a private document each counted as probatio semi-plena (half-proof); only two eyewitnesses or a confession made probatio plena (full proof) (claim-roman-canon-law-rated-one-witness-equal-to-a-private-document). Crucially, the system permitted addition: two half-proofs on the same point could combine into full proof — "except in criminal cases, in which proof must be clearer than the noonday sun," and likewise in matrimonial and grave civil actions (claim-two-half-proofs-combined-to-full-proof-except-in-grave-criminal-cases, already flagged [unverified -- needs primary] for the exact three-way clause, independently corroborated in spirit by a wholly separate French secular tradition per that note's own audit trail). That note already draws the vault-mirror explicitly, in its own words, written before independent_roots existed: "a load-bearing quantitative or mechanism claim may not be hardened by piling up two weak corroborations." The historical system, examined closely, has the same shape the vault's own mechanism turns out to have — a graduated count, with a hard categorical refusal reserved for specific, named, high-stakes chokepoints — rather than the flatter "unus testis, nullus testis" floor that the Latin tag and Note A's "two witnesses" phrasing most readily suggest.

Claim: the resemblance between the two target notes is real, and it is a mechanism-level partial correction, not a confirmation — "graduated bookkeeping with narrow categorical chokepoints," shared by both systems, is a tighter fit than Note A's "run in reverse" image of a flat revived bar

Claim type: synthesis across the above (historical + technical-mechanism; floor met by the component claims above, each individually sourced to its own appropriate tier).

Put together, the three findings above resolve the pair as a genuine mechanism match that complicates, rather than simply confirms, Note A's framing. Note A's prose ("the vault explicitly refuses... reimposes fixed proof-rules... the same apparatus, opposite direction of trust") most naturally evokes a flat, general, always-on bar — the popular reading of unus testis, nullus testis. Note B's close reading of the actual mechanism shows something narrower: a count gated at two specific checkpoints. And a closer look at the historical side of Note A's own comparison shows it, too, was never simply a flat bar — it was a graduated system with its own narrow, named refusal-points (the criminal/matrimonial/grave-civil exception to combination). The genuine connective tissue between the two notes is not a shared word (they share none) but a shared shape: both the medieval apparatus and the vault's own apparatus turn out, on close reading, to be graduated accounting systems that reserve categorical refusal for a small number of named, high-stakes chokepoints, rather than blanket prohibitions on single-witness or single-root evidence. Note A's "run in reverse" headline survives as directionally right — the vault did reimpose a graded-proof regime where the Enlightenment abolished one — but its specific claim that the vault "explicitly refuses" single-witness corroboration, read as a flat categorical bar, is the part this capture's evidence narrows: the refusal is real but local, in both systems, not general in either.

Further leads

• claim-poullain-du-parc-two-half-proofs-cannot-make-a-full-proof and claim-clearer-than-the-midday-sun-standard-attested-independently-of-catholic-encyclopedia — the independent (French secular, Tier 1) corroboration of the non-additive maxim's spirit, not yet connected explicitly to independent_roots in any existing note.
• claim-bloemberg-frames-preuve-legale-abolition-as-jointly-epistemic-and-political — Bloemberg's own anti-monocausal argument (epistemic change and distrust of professional judges) is itself a small instance of "don't let one story stand in for the whole explanation," unexamined here as a further mirror of the vault's own floor.
• observation-suspicious-perfection-independence-absence-signals-defect — the sibling observation that independence-absence shows up as suspiciously good agreement, not just thin sourcing; not cross-checked here against the chokepoint framing above.
• Whether root_shared as currently specified could itself be read against the medieval "two half-proofs from the same ancestor" problem — i.e., whether the vault's bookkeeping has a combination rule of its own, analogous to the criminal-case exception — is an open structural question this capture raises but does not answer.

Entity candidates

• Constantine I — person — the AD 334 constitution behind the entire unus testis lineage that Note A's "two witnesses" phrasing gestures at; already flagged once before (2026-10-19 capture) as the foundational figure the comparison is measured against. Already has entity-constantine-i — flagged again here only because this is the second capture in three months to lean on his rule without naming him in its own prose; worth checking whether his page already carries the half-proof/combination-rule nuance this capture adds.
• Azo of Bologna — person — the 1190s Glossator credited with the half-proof examples (single witness, private document) that give the medieval system its graduated shape; already has entity-azo. Flagged first per the standing blind-spot instruction: he is the earlier figure the "flat two-witness bar" reading of Note A should really be checked against, not a co-author of either target note.
• R. G. (Ronnie) Bloemberg — person — already has an entity page per the retrieval list (entity-r-g-bloemberg); his AJLH article is the Tier-2 backbone for the half-proof/full-proof definitions this capture leans on. No new page action needed; noted for completeness.
• Adhémar Esmein — person — no existing entity page found. The French legal historian whose 1913 compiled volume (via R. Garraud's appendix) carries the independently-corroborating Poullain du Parc maxim cited in the half-proof cluster; a bibliographic hinge the vault has cited repeatedly (per the 2026-08-08 attribution-fix audit trail on claim-two-half-proofs-combined-to-full-proof-except-in-grave-criminal-cases) without yet giving him his own page.