Research: audit #2979 profile against kappa=0.1

[Gonzalo-Mier]

This research package independently audits the profile-bearing stages and T74 owner allocation for the #2979 candidate, then checks the donor's κ=0.1 target against that profile and the newer six-skip profile from #3042.

The exact target screen uses κ = a/(1+a), so κ = 1/10 requires a = 1/9. At that point, exact integer ninth-root bounds show that the paid local children alone contribute at least:

• Conditional κ ≈ 0.0017290339826465: joint rank-7 end frames, five paired extractions and #2957's rewrites on #2942's words (+5.00e-6 over #2969) #2979: 48304277901/44800000000 > 1.
• Conditional κ = 0.001729736166387916: six restore carrier/pivot skips (R1_E42, R2_E7) on PR2979's words and literal allocation #3042 six-skip profile: 48295445607/44800000000 > 1.

These floors omit idle children and centre costs. They rule out κ ≥ 0.1 for these fixed local profiles under the inherited moment and transfer, even if those omitted costs were free. They do not rule out a changed local profile, coverage law, or outer transfer. The selected frontier is unchanged and this PR claims no κ improvement.

The verifier reconstructs the V45 coordinated lowering and losing single-helper control, replays the profile-bearing extraction/rewrite stages for both providers, checks 1,332 owner-child windows and all 271,728 owner roles, verifies rank-mass 4,816,512 and reconciles 66,813 physical families. It also pins the PR #3042 claim used by the six-skip screen.

Validation passed:

python3.12 -B scripts/research_package.py --package research/rank7-v45-profile-audit
python3.12 scripts/check_research_commit.py --package research/rank7-v45-profile-audit

The audit does not independently certify extraction/rewrite operator identities, compensation and cleanup semantics, global stock/precision/native admission, or outer release. Please retain attribution to #2979/#2957 and the six-skip PR #3042 lineage; the package discloses substantial OpenAI Codex assistance.