Every section banner dies (103 -> 0) across test/lang, the observer
suites, the C carriers, and the five comment-heavy corpus fixtures;
banner provenance (#N cites, carrier numbers, repair-cluster labels)
folded into headers or adjacent WHY comments. Narration deleted; row
provenance, ref cites, divergence pins, and layout contracts kept
(fwd-ref decl-order guards and bootstrap-gate corpus rationale
restored where the sweep over-cut). Comment-only proven: all 3742
wwbuild workdir .s byte-identical before/after; test-commit and
test-byteid (161 lang + 1399 data, 0 pinned-divergent) green.
m3sep_test.ww drives the raw w6c/w6a/w6l toolchains (+_ww twins) over
hand-composed //ww:module units splicing produced .wwi bytes: the
bodies==.wwi keystone on both stages per package (mid/root/smid) with
determinism, rule-10 and the value-global DATA guard; the end-to-end
i32 chain (exit 59) with final-exe byte-id; and the #49 str-label /
#53 export-leaf link legs (exits 160/140) with their .s needle
proofs. Each @test re-produces its .wwi cold instead of sharing the
carrier's single scratch dir (same claims, sharper isolation).
