Certify The Velveteen Rabbit from frozen Phi

[dcellison]

Summary

• Certify The Velveteen Rabbit as the eighth translation completed under the isolated D102 process.
• Freeze 426 aligned Phi units at 16597a3af79e10b9ff4e66638108da09e4c9fb6bce6ccd4d275ccb4b841c3f59.
• Preserve all 429 Williams citations, which reconstruct 20,296 normalized source characters exactly and in order.
• Publish the final aligned layer at 42ab9396fd582eb5a3d6970eaea221f80a6fb5c0707bfd351121095c15e272c5.

Translation

The source-to-Phi pass changed 251 final units. It gives six recurring figures stable Phi onyms, puts ordinary temporal and locative adjuncts after explicit subjects, and repairs tense, participant roles, purpose clauses, embedded questions, and complement scope. The last focused repairs cover the lined ears, Timothi's identity, Tupiwa's boast and thought, the moving fern, the healer's instructions, and the different bedroom. The translation adds no root, module membership, registered compound, function word, or grammar rule.

The Williams source, previous English, notes, and limits remained hidden during the source-to-Phi phase. After the Phi stream was settled, a fresh non-forked context received only the anonymous Phi packet, canon, grammar, lexicon, and voice references. It derived every exact gloss and English reading from that frozen Phi. Independent source-blind review exposed further Phi faults; the affected English was discarded, the Phi was repaired, and fresh contexts derived and audited those units again. The two fidelity directions remained separate throughout, and the source was never compared directly with the derived English.

The reader-facing notes now describe the finished translation rather than its editing history. A separate Humanizer and Phi voice audit removed contract-like process language while preserving every Williams quotation, Phi sentence, and exact gloss.

Site and records

• Mark D111 complete in the machine-readable certification register and regenerate its readable ledger.
• Refresh the roadmap, corpus review, development log, and handoff state with The Prophet, On Love next in the queue.
• Teach the site renderer to recognize a valid productive name-form after ne, which allows ne timothi to render without treating it as a lexicon root.
• Update the editorial stanza count from 427 to 426 after two Williams fragments become one aligned Phi unit.
• Regenerate phonetic-neighbour attestations against the final corpus.

Checks

• python3 scripts/validate_examples.py: 1,301 entries, 0 errors, 0 warnings
• python3 scripts/validate_sentences.py: 3,306 examples, 0 errors
• python3 scripts/validate_sentences.py --paths texts/velveteen_rabbit.md: 1,839 examples, 0 errors
• Williams reconstruction: 429 citations, exact and in order
• Detailed versus continuous Phi: 426 units across 18 sections, exact match
• python3 scripts/translation_process_status.py --check: 8 certified, 12 pending
• Sentence, vocabulary schema, content-decision, name-form, translation-layer, and certification-register test suites
• Generated Part VII reference comparison
• Phonetic-neighbour baseline regeneration
• python3 scripts/build_site.py
• git diff --check

The full --docs sentence scan parses 5,523 examples and retains its two recorded migration diagnostics in documents/evaluation/narrative_test_corpus.md, PHS103 at line 46 and PHS115 at line 212. This pull request does not alter either case.