Research: execute complete generic packets with nested child dispatch

[jmbarrancoml]

I replace the address-oracle child boundary in a generic caller with completed digit-swap/ADD/sign leaves or a smaller complete caller. I connect those operations to actual full-ring preparations and generated weights.

I prove the embedded-field interpretation: collapse the complete fields before, between and after two child chunks, retain the full high prefix, and preserve the exact serialized volume without moving records. The executable backend uses two complementary projector residuals in dimension two. It gives a closed caller for testing this interface, with no exponent-saving claim.

I checked 1,456,542 distinct 32-bit records at the first residual and final weighted endpoint. For both Z/9 branch fixtures, I retained all 81 high matrices and all 9^4 body addresses. I also executed a nested unit-reset child through all 48 fresh GL2 low classes, starting from an actual 64-dimensional preparation, and checked restoration of a buried nonzero caller frame. I accounted for 872 children: 871 completed leaves and one internal caller. Six controls exercise low-only coefficients, missing reversal, lost global slots, incomplete packets, insufficient stock and invalid chunk geometry.

I passed the registered normal and optimized Python 3.12 replays. Reproduce them with:

python3.12 scripts/research_package.py --package research/selected-joined-dispatch

Review PROOF.md sections 1–5 and the boundary in engine.py. I implement payload copies and parking through sequential record macros and charge their full bit volume. I leave host arithmetic, memoized immutable preparations and control objects outside the measured payload machine. Reviewers still need a selected GL64 family/owner/hosting execution and a full binary transition-table compiler before using this contribution in a multiplication-exponent claim. I leave the selected frontier unchanged.

Concurrent PR2332 by @commdata2338 connects PR2234's actual selected matrix preparation to PR2310's control tables. I contribute complementary complete generic payload and child execution. @aakashthakkar and @thakrarsagar address borrowing-stock fit in PR2324 and PR2343, under their stated caller contracts. Their large-prime bounds do not certify these q=3 fixtures.

The next step is to join the dispatcher to the source-bound selected family/owner and XOR/copy/erase interpreter, with a checked control-tape encoding. TASK.json records the handoff. Coordination: Discussion2319.

Jose Manuel (jmbarrancoml), with substantial OpenAI Codex assistance. I retain credit for source527, commdata2338's normalizer/descriptor and parking work, Ross Richardson's fallback algebra, the selected supplier lineage and the preceding stream/control implementations. I claim no independent human or new formal verification.