feat(adr-0246): slice 0 — benign nominal-frame mismatch diagnostic (evidence-only)
Diagnostic-only first slice of the ADR-0246 programme: classify WHY benign
trajectories mismatch the declared identity frame, using the preflight brief
§3 instruments implemented eval-side (induced action A(F), d_orth, d_stab vs
locked H_id={I}, typed e4/e5/unclassified residual channels, plane occupancy).
No gate, threshold, axis, flag, or corrector changes; serving untouched.
Verdict (docs/audit/adr-0246-slice0-mismatch-diagnostic-2026-07-17.md):
all 25 live benign/paraphrase turns = foreign_leakage; precision transport
immaterial (<=3.6e-5); path accumulation ruled out (per-turn d_stab already
0.15-813); declared frame statistically unspecial vs 32 random control frames
=> semantic coupling absent, confirming + sharpening the D4 root cause.
Two benign sub-populations resolved: 18/25 boost-involved (e5, non-isometric),
7/25 pure e4 conformal tilts (near-isometric).
[Verification]: uv run core test --suite smoke -q => 176 passed;
tests/test_adr_0246_mismatch_diagnostic.py 16 passed; adjacent identity
surfaces (gamma_calibration, identity_manifold, identity_gate wave/runtime/
eval) 75 passed. Gate surface pinned untouched (flag default-off,
_WAVE_LEAKAGE_BOUND unchanged).
b9c54e7713