DONE: complementary B/C cancellation is exact; fixed-owner serial consumers lose

[Clear1nsight]

Status: DONE / 2026-10-22T21:55:55.157Z / priority revision 7.

PR #4848, exact head 890fadd4683ac206fa0c39ad58ff80932bdba26b, gives an exact typed C→B complementary cancellation and a scoped interval-wide serial-consumer ceiling. The pair retains all 1,040 independent inputs and helper jobs; its ordinary ports require 56 genuine rank-one query children. Complete paid/target rank masses are 9,579/9,403, a deficit of 176. This is a typed-interface construction and written cost proof, with finite exact corroboration. It is not a fresh complete native trace, profitable caller or new kappa.

For n>=2 whole calls on the same 224 main owners, one shared central Q, n disjoint seven-axis blocks and only the source helpers as additional productive fields, all main target ranks are capped by 7n+1. The complete optimistic moment is strictly negative for every 0<C<1, even granting free cleanup and finite work. Two C calls only tie the mass bound at C=0. Changed target ownership, genuinely shared producer work or a different paid profile are outside this ceiling.

Base 624de9dc66858af0945e5f8a0942488d62280d43. Exact #4703/#4712/#4674 pins and unchanged source snapshots are in the package. The native/phase audit is inherited from #4712; #4674's direct-price method is credited. This changes the consumer and uses the refreshed C profile; it does not repeat the autonomous-B/A-path or 120-head-query rejections.

Local registered replay, committed-byte/source-history preflight, Git-free optimized replay and actual receipt/source corruption checks pass. A second agent independently rederived both derivative bounds and the universal-n inequality, recounted the profiles and checked a rational non-involutive-Q fixture and raw dual. Parent also replayed the scientific verifier. No human or proof-kernel acceptance, new native phase suite, precision/compiler/outer admission or selected cold replay is claimed. The original hosted run failed before scientific replay due to a Git maintenance fixture-cleanup race. Focused repair #4857 is now included at additive head d5772576a9f260768d39eea0adde36e08602e889; its 16 scientific files retain exact original bytes. Hosted run 39863812767 passes all 347 helper tests (11 expected skips), the scientific package replay and committed-input mutation gate.

Smallest useful next task: stop adding calls inside this losing serial model. Propose one actual shared producer or changed ownership contract, give its disjoint target census and optimistic complete cost bound, and proceed to emission only if that bound can change the decision. The ongoing actual-centre chronology work is tracked separately in #4840, and block-unit physical realization in #4843. No active reservation remains on this completed serial model.

Reproduce: python3 -I -O -B research/complementary-consumer-feasibility/verify.py (small standard-library run), plus the registered package and committed-byte commands in the PR. Credit Dugongue, Cao Yifan/intqwq, Cinna Davis, zxv, Jacob Alan Sussman, DaysSky and all retained source authors. Chris R / Clear1nsight, substantial OpenAI Codex assistance with delegated construction and separate agent verification; Apache-2.0. Live overlap was bounded and cannot certify novelty.

{"schema":"integer-mult-coordination/v1","kind":"DONE","task":"complex-new-helper-inventory","scope":"Exact refreshed PR4703 complementary pair and all-exponent serial fixed224-owner consumer ceiling","base":"624de9dc66858af0945e5f8a0942488d62280d43","expiry":null,"artifact":"https://github.com/CrocSwap/integer-mult-bounds/pull/4848"}