[petalas]

Summary

The selected affine payload interface cannot contain PR2224's regular S3 convolution on twelve independent data arrays through affine input/output maps.

selected fixed linear operations + affine boundaries: mixed difference = 0
regular S3 convolution: mixed difference(e_g,e_h) = e_(gh)

This closes only the literal affine-data placement question. It confirms PR2224's rank-nine tensor identity and leaves fixed-coefficient, operator-valued and nonlinear replacements open. Tensor rank nine is not established as one recursive address-rank-nine child. No new kappa, supplier or selected-result change is claimed.

Evidence

• Before: PR2224's favorable eight-block histogram had no actual selected source map.
  After: both frozen 24,767-event orientations are classified; all 36 exact tensor products have nonzero mixed differences. A written affine-invariance proof rules out the stated placement for arbitrary dirty helpers.
• Four semantic controls and 36 fixed-coefficient positive controls pass. Registered Python 3.12 normal/optimized commands and an optimized Python 3.15 replay pass in the exact committed regular-only, inode-disjoint read-only Git-free export with TMPDIR unset. Private source-byte and Boolean-receipt mutants reject.
• Selected native linear semantics remain inherited. No native endpoint, owner, precision, compiler or outer replay is claimed. The changed physical bill remains unknown. Exact sources, proof, limits and Open next questions are in research/s3-selected-affine-boundary/.
• Designated worker2 scoped peer review found no blocking standards/spec findings. Its verbatim hash-bound receipt is PEER-REVIEW.json; it does not admit a supplier or the full multiplication theorem.

Merge Danger

Door: two-way

Blast Radius: isolated

Adds one source/proof/reproduction package. Predecessor sources, selected science, shared code and workflows remain unchanged.