Research exact local P104 I-G/2 tail correction

[1337isnot1337]

Result

This adds an exact finite rowwise I-G/2 tail correction to the pinned P104 low-rank G/2 channel. At the cut after the 4,058-event channel prefix, all 104 X source registers are full-frame and each of the 104 Y targets is in its rank-eight endpoint frame. For each target, the word promotes to the full frame, adds the 48 nonzero entries of (2I-G)/2, then restores the endpoint frame.

The registered package proves the exact arbitrary-input identity over all 280 formal columns and checks 208 native rank-one frame-change recipes. It records 4,992 scalar additions, four mutation controls, and deterministic witness/recipe hashes.

Local cost and limits

The augmented local paid profile is {1:1282, 2:58, 3:86, 4:104, 5:48, 6:24, 7:8, 8:8}, rank mass 2,576. Under the same exact local relaxation, its ceiling is lower than the frozen uncorrected channel by the rational interval in RESULT.json; the relaxed moment at a=1/99 is greater than one.

This is a partial local construction, not a complete supplier or a new kappa. Global owner/host scheduling and copied-information delivery for the source reads remain unproved, and the word is not integrated with the chronological delivery schedule in #1940. The channel and native realization in #1972/#1991 remain frozen and unchanged.

Provenance and reproduction

• Seed 749717109203f916f620456c85e26673, complex-mechanisms, task p104-side-correction, priority revision 4.
• Scientific base 7c1e7811c730b17fe0eeef6c82ef011318b0465e; selected-record SHA256 10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32.
• The package records exact source hashes and the 2026-10-14 upstream/overlap checkpoint in TASK.json.

On Python 3.12, from repository root:

python3 scripts/research_package.py --package research/p104-tail-side-correction-seed749717 --check-only
python3 scripts/research_package.py --package research/p104-tail-side-correction-seed749717

Both commands passed after rebasing this branch onto current main. Substantial OpenAI Codex assistance was used for verifier implementation, exact screening, and package preparation; disclosure is included in TASK.json.