Clarify sufficient reuse and complexity review (6.1.2)

[mzored]

Simplification now compares solutions that preserve the complete accepted behavior by total cost. A short diff, native feature, or existing dependency must meet the actual constraints. Complexity findings name the redundant construct, consequence, maintained replacement or justified deletion, and responsibilities to preserve.

The original regression corpus adds a shared-invariant repair fixture and paired native-capability recommendation probes, and strengthens the maintained-composition review case. The balance oracle accepts two complete repair shapes and rejects the unfixed and transfer-only implementations. Kernel, workflows, authority, public skills, package structure, and installation remain unchanged. Version, manifests, evaluation metadata, evidence records, site evidence, and changelog align at 6.1.2.

Validation:

• Focused oracle/corpus/preflight checks: 64 passed.
• One full local package gate supplied the static signals. Its pytest signal initially failed because the verification wrapper placed temporary fixtures inside the checked source tree. Rerunning that affected signal with owned scratch outside the checkout passed all 396 tests on the unchanged final candidate tree 40b940906d99373f3c16a7a2dd7dd5f694037e4b (pinned dependencies, Python 3.14.7).
• git diff --check origin/main: PASS.
• Isolated read-only Codex review: APPROVE, no qualifying findings or wording preferences; both host and OS homes isolated, minimum auth referenced rather than copied and removed on exit.

All new and strengthened behavioral cases remain UNVERIFIED. No paid behavioral comparison, measured savings, or improved model reliability is claimed. Prose changes require no new clean-install experiment; host schema, install, activation, continuity, and behavior stay separate in the release matrix. Publication is owner-authorized; tag the integrated revision only after main CI passes.