[petalas]

Summary

The actual-width T74 proposal leaves a sparse-axis exposure router unpriced. This package proves a sharp obstruction for implementing that specific endpoint through completed B1 calls:

one contiguous equal-block exchange: at most 4 new broken adjacencies
g isolated sources -> g disjoint isolated targets, support <= 2g: exactly 4g
minimum over allowed extensions: exactly g completed B1 calls

Arbitrary intermediate spectator movement is allowed. Each complete owned packet return costs at least V writes, so the literal whole-packet adapter has tariff Theta_q(Vg). The hard family meets the proposal's exact spacing and width domain. A fused router, larger-support extension or changed consumer/view contract remains open. No supplier, integer-kappa improvement or selected-frontier change is claimed.

Only research/isolated-axis-exposure/ changes. The proof is self-contained; its matching physical upper bill retains PR2899's completed B1 premises. Canonical source bytes are pinned; pending T74 work is motivation/comparison, not a hidden Git dependency. Credit stays with Jose Manuel, David Arias and the retained ordinary/native/compiler lineage.

Evidence

• Before: the actual-width handoff explicitly leaves arbitrary-support routing as a conditional interface. After: the all-size four-cut proof excludes a sublinear support power for the stated black-box B1 adapter and identifies which contract must change.
• CPython 3.12.15 normal/optimized registered replay passes. Exact checks cover 188,948 exchange comparisons, all 4,365 supported completions in 30 families, and independent all-shorter-word censuses for g=1,2,3. Six actual optimized CLI corruption/missing-evidence controls pass. Adjacent support and larger endpoint support give explicit cheaper controls.
• Final publication byte/source-history checks and actual committed Git-free scientific/control commands pass. The export manifest and before/after source/export identities are retained. No production GL64/T74 payload, independent human/formal acceptance or full selected cold replay is claimed.

python3.12 -B scripts/research_package.py --package research/isolated-axis-exposure
python3.12 -B scripts/check_research_commit.py --package research/isolated-axis-exposure --check-source-history

Canonical investigation. Main and priority revision were refreshed before publication; the scientific baseline is canonical 20ecf2891d80b5acf18d980b2164fb44ea7faa90, revision 6.

Merge Danger

Door: Two-way

An isolated research package; predecessor inputs and shared defaults are unchanged.

Blast Radius: Package