Prove sharp selected-centre repair dimension for owner blocks up to six

[RossRichardson]

Permitting each retained repair row to mix a block of up to six physical cover owners does not reduce the exact 104*b additional scalar-linear-coordinate requirement for the selected PR #1094 calls 0/4 observation model. A written joint-kernel support argument uses exact rank-eight/column-pair premises and a new certificate: all 12,432 ordered distinct-port route products are unique. Seven owners are only a necessary support threshold, with no existence claim at seven.

The package retains the entire-later-query requirement, all old query observations, disjoint repair blocks and arbitrary global nonlinear decoding. It builds on Aaron Wang/navinor's PR #2391 and ywz65's owner-local repair theorem, with their native/source and selected-emitter lineage credited. A separate reverse-adjoint and chart-swap implementation checks the finite premises. Scope, proof, exact sources and next questions are in the package README; coordination is in Discussion #3062.

Validation passed the complete registered replay on CPython 3.12.12, optimized primary replay on CPython 3.14.6, six primary mathematical controls, three independent premise controls, three behavioral integrity tests and the exact committed-file/input check. Reproduce with python3.12 -B scripts/research_package.py --package research/centre-block-repair-information.

This is a scoped information obstruction under the retained constant-payload/native-interface premises. Coordinate dimension is not a recursive tariff. No native recovery program, complete recursive-cost rejection, formal-kernel proof, new kappa or selected-frontier change is claimed. Ross Richardson, with substantial OpenAI Codex assistance (GPT-6.1 Sol).