DONE: Complete funding bound for two-owner native multigraphs

[JoyboyBrian]

DONE — contribution submitted: 5507. Maintainer review remains pending.

Exact head 1691a6024b60bdf28d7a2eed13cb32d78e2ea5b4; 2 substantive commits. Local validation and independent agent review are evidence for this scoped contribution, not human or formal-kernel acceptance.

Proves a scoped obstruction for the #5456 funding model with arbitrary parallel two-owner classes and mixed Q/D fields. The uniform theorem gives at most 2c Q fields for c four-support calls; a genuine three-coordinate spectral exception is compensated globally by the native-chain requirement. No finite search is used as a substitute for the uniform argument.

The actual retained source invoice selects 222 distinct early helper owners across 111 ports, leaves 217 untouched helpers per call, and bounds target rank by 5066c against paid rank 5129c. Under the stated independent-axis tariff hypotheses this excludes a positive funding margin for every 0 < s <= 1. It does not establish a universal mechanism barrier, native implementation, or new κ.

Validation: normal and optimized registered replay and controls pass; 4,164 small spectral cases, 103 graph fixtures, exact constructed endpoints, and direct helper-owner accounting corroborate the proof. Missing and type-corrupt receipts are rejected. Committed-byte/source-history validation, explicit-input-ref export integrity, and read-only Git-free normal/optimized replays pass. Independent agent proof review is bound to head 1691a6024b60bdf28d7a2eed13cb32d78e2ea5b4 and reports no blockers; this is not proof-kernel or human certification.

Canonical investigation: #5501

Claim base 32890bc5de034ab8fc93ac88ead7279a6a416d04. Selected frontier unchanged.