[petalas]

Summary

Independently certify dcharb78/PR3182's universal fixed-cover P104 exclusion, including same-layer source acquisition and early target writes.

three pinned JSON inputs
  reconstruct all 2,309 compatible endpoint pairs
  read sources before each layer's contact step
  compare interval contacts with literal binary frames
  accumulate every target write
  retain exactly one missing coefficient: C[71,102] = -1/2

Under the stated immutable-X, unread-Y, independent-dirt, monotone equal-frame contract, every actual same-dimension dependency run collapses to one direct contact. Every matching embeds into the pooled test. The dimension-five forward cone has 1,725 edges and the dimension-six backward cone has 37, with no intersection; both early Y71 writers also lack X102.

The universal cut and its numerical values belong to PR3182. This adds the support-inclusion proof, an implementation importing no predecessor executable/results, literal-frame relation comparisons and rejecting controls. PR3214's output/cycle audit and PR3252's distinct memory theorem retain their credit. The original channel's G/2 identity remains an inherited premise.

Evidence

• Before: PR3182's strongest universal exclusion awaited this independent proof/replay after separate output/cycle claims needed correction.
  After: both complete contact engines reproduce all 104 output masks, all eight relation hashes and the sole missing entry among 4,992 requirements. Seven behavioral groups pass, including exhaustive F2^3 interval checks, all 4,096 actual three-frame assignments, fresh reads, multiple sources and source/geometry corruption.
• Registered CPython 3.12.15 replay with optimization and warnings as errors passes. A read-only Git-free CPython 3.15 export passes literal replay and controls. Strict Ruff diagnostics and metadata/source pins pass.
• A separate read-only peer check inspected the proof/source/pins and independently passed both full modes and controls against the unchanged eight-file reviewed diff 69f4593253ec00ba598e1be733931e550606295af6d2707f858dc52e5fc1f9cf.
• Replay: python3.12 -B scripts/research_package.py --package research/p104-pooled-cut-independent.
• Base main: 463113d675af7410ac91d53a4fed51ec6494eb9a, priority revision 6. All live inputs use canonical main Git blobs; no pending-head Git-object dependency.
• Canonical investigation. Source pins, exact receipt, limits and Open next questions are in the package.

This is a scoped proof and finite certificate. The other 4,991 reachable requirements are only a necessary test. No chronological coefficient program, arbitrary dirty restoration, native/physical/precision/consumer invoice, all-size theorem, general lifetime floor, new supplier or kappa is claimed. Selected cold replay, original/native G/2 proof replay, formal-kernel and independent human review are unrun. Selected and frozen predecessor files are unchanged.

Merge Danger

Door: two-way

Blast Radius: isolated

Adds one opt-in research package and its declared replay. It changes no supplier, selected result or shared implementation.