[aliromman]

The fixed native schedule misses helper cuts that become possible after a paid intermediate meeting. This construction splits existing queries through common frames, then retires 23 independent helper inputs and supplies 23 independent continuation inputs. Eighteen cuts use one subdivision each; a three-channel meeting shares five subdivisions, and two further meetings use two each. All 27 added queries per local call are charged.

The complete conditional exponent increases from PR #5718's 0.00311024570 to 0.00311301600: +0.00000277030, or +0.0890701%. The exact raw crossing is (0.00311301604454, 0.00311301604455). This is below the donor's 1% target. The complete p16/a7 caller has 13,390 productive targets and 66,834 paid calls, with rank surplus 760. The extra targets require cutoff 16, floor 17, tail offset 1/7 and decay 125/128; the old completion bound is explicitly rejected. Scalar, route, owned, normalization, idle and compiler costs remain funded.

The focused package is research/p104-paid-meeting-splits. It contains the exact constructor and plan, proof, authenticated PR #5718 source export, retained attribution, full result and rejection controls. Validation covers all 776 public/dirty-owned columns, 2,920 native maps with exact phases, the complete physical caller, two rational interval engines, the compiler, 54 strict outer inequalities and four functional checks. Run:

python3.12 -B scripts/research_package.py --package research/p104-paid-meeting-splits

Publication commit: 8483f9a679d06c1e270141dbc686a7e76e7a7522. Registered normal/optimized replay, metadata/source pins, committed bytes and executable modes pass. The official export has 20 package files and SHA256 5a3046462b06c108e2853898c9daf4b466bc46f47e5f77983515b6e5e7ed5023; export integrity passes. The separate isolated optimized replay passed in 165.589 seconds with a read-only source tree, empty PATH and no Git directory; every exported byte remained unchanged. The full-history flag is not run in the shallow checkout; both declared source commits pass direct ancestry checks. VALIDATION.md records the scope. Hosted CI failed in the shared planning tests before package replay. test_research_current_assignment.py still expects the legacy command mappings after priority revision 8. Its eight failures and two errors reproduce locally; that test, both picker scripts and the catalogue are byte-identical to main. The scientific replay job was skipped. This is not a hosted scientific pass. Independent acceptance remains pending.

Constructed from canonical main 553d26d796fde9871ff2661a49ac7732e945410f, with PR #5718 pinned at a1534ac45401c9a9c80c07aa65c89e00e33594f8. The publication commit is rebased onto the final upstream checkpoint at de5ec2e7a6074068452f45b1a191b5c72832df4a was read, including priority revision 8 and its complete helper-inventory gate. The frozen research sources are retained. The canonical Discussion records the claim and findings. A later overlapping investigation released the shared scope and published a 17-split checkpoint without a duplicate PR. Its nonnested-excursion work remains separate. No global optimality or exclusive discovery claim is made.

Credit flyingviper's PR #5718, Ali Romman's PR #5713 and complete P104 caller, Joel Pulikkan/GamingPuzzled's PR #5684, the PR #5599/#5609 helper interface, the published PR #5680 cut checks, and every retained native/controller/recursion/compiler/BIT source. New work: Ali Romman with substantial OpenAI Codex assistance; no subagents. Apache-2.0.

The five inherited framework assumptions remain. The huge uniform word and complete rounded multiplier are not materialized. The selected frontier is unchanged; no merge or closure is requested.