PLAN v1 response

The fresh Opus review requested changes at 2026-10-13 01:41:16 UTC, raw SHA256 c253af5655b79f1fbdfa4fae0cd4baf6cd2ae42d741b63e178be6aa496bd3b94. The complete v1 packet and review remain unchanged. No existing audit artifact or production file has been edited. PLAN v2 addresses all six blockers:

1. Measure both ce0 and candidate in one session through the same local Git URL, with separate fresh caches/consumers. Check the external nostd-parent manifest before, between and after; require the ce0 warning and candidate absence.
2. Specify a child nu -c whose code contains a quoted literal source path, then calls the original private prepare-crate. That mechanism has now run on both historical revisions: all 16/17 exported cyber-bao files and three copied harness inputs match, allowing only the original function's exact Hemera substitution. Retain the full child code, raw receipt and all five surrounding capsule inputs.
3. Limit byte equality to reader stdout and unchanged production source/manifests/ locks. Enumerate the exact Radio revision, consumer ID and physical-path metadata substitutions; preserve every other field. Historical HTTPS metadata is evidence, not the new local-Git pair's equality baseline.
4. Add 19 machine-readable historical index bindings, including repeated input indices, two old PLAN context bindings and the already historical pre-CODE verifier variant. They identify each satisfying archive and intentional physical mismatch. Original indices and historical hashes remain untouched.
5. State defaults per script. The two verifiers resolve from the original capsule; the reproducer preserves ~/cyber/radio and ~/cyber/hemera, even when invoked from another worktree. Both historical HEAD assertions remain; only research's existing verifier checks tracked cleanliness. No new clean-tree assertion is added.
6. Rename the outer copies to historical-ce0-reader-before-runs.json and historical-ce0-reader-after-runs.json, preserving bytes and timestamps. They bracket the historical reader runs, not the packaging change. V1's own frozen copies retain their original labels and bytes; future runs get arm-specific names.

Optional advice: adopt an audit-basename guard and document moving, rather than duplicating, restored .txt files inside temporary layouts. Keep the literal production-tree equality check. Decline global source-manifest cleanup: it expands this task without evidence of another consumer defect. The historical nostd-parent absolute path is nonportable but reconstructable; calling it permanently unreplayable would overstate the limitation. Its bytes remain unchanged.