[GoGoKo699]

The aligned rank-one construction exposes a lookup-dependent phase, but its apparent simplicity does not reduce the complete-indicator problem. This change proves resource-preserving reductions in both directions and tests a broader accepted-source model beyond the Bell envelope.

• A general complete Pauli-derivative source, including nondiagonal address/work-moving representatives, produces the indicator with two source calls and the same derivative error. Conversely, two indicator calls and a paid four-T-band shared-control bank implement the displayed phase with twice the query error and only returned dirty work. The fixed-accuracy count, width, gate and sublogarithmic-depth targets are equivalent. The existing indicator count theorem gives T >= m/2 for the derivative source below error 1/sqrt(2); it supplies no growing depth bound.
• For any net address-controlled Clifford source, Gram error strictly below 1/4 for (I+Q_x)/4 forces that Gram exactly. One clean signal has error floor 1/2. In the separate read-only affine-XOR query grammar, arbitrary interleaved fixed Clifford operations still require k(k+1)q >= p; with two clean signals, 3q >= p. This is a query-word restriction, not a native T-depth lower bound.
• Update the existing proof homes, evidence/route catalogues and 220-line checkpoint. The next question is a complete source using genuine non-Clifford dynamics, changed address motion or joint native synthesis. R96 and all general frontier claims remain unchanged.

Starting main: a7a7cd213cf6df366a4c4e26261628caa645b03c.
Reviewed head: d8831797a750dd20f93cab86e5841e1cbefd78a3.
Reviewed tree: 6f5154990788c165e73b69841b9dde1c87e5c4ce.

Independent internal proof audits and ten new exact tests passed. Navigation/notation checks passed. Focused rendered review of both proof documents plus the presentation guide passed at desktop and narrow widths (3 documents, 5 diagrams, zero failures), with visual inspection. Report SHA256: 931a2829a66f0415bc4b2db2be288794f6c88eb6bcc3c72ca06105020c4af93d.

The complete ten-command local gate passed on the reviewed tree (Python 3.12.14 / NumPy 2.3.5; 1,092.526 seconds). The captured test summary is 1,741 tests, OK. Every command returned zero and the tree remained unchanged.

• Gate receipt SHA256: 7450d4d392edd4efae585e4c35d253257eeacb03a0a2c6ac8bda1acd6757664f.
• Complete gate log SHA256: cb7c1845ca61aa0248295e5ebc883b115e0e7df5e8f5e617d2f8e78ed66ea217.
• Full presentation CI passed. Artifact 12584153071 SHA256: c808d0c6a035ab4eada04072eb14764af5a36ac9eec7bcbc9b52dcd2fe21067d.
• PR validation and push validation cover Python 3.11 and 3.13 on the same head. Both runs passed both Python jobs; all three CI runs and all five jobs completed successfully on the reviewed head.

Internal proof audits and finite fixtures support the stated analytic results; they do not establish a new unrestricted native-depth bound.

Merged as 39d4c4c436d9751f7b80aec9eb92a51b7ebd3f73. The merged main tree is exactly the reviewed tree 6f5154990788c165e73b69841b9dde1c87e5c4ce; the local main checkout is synchronized and clean.