Repair the Week 4 learner product spine

[skyzh]

Summary

• establish one learner-visible product spine across the nine Week 4 mechanisms
• describe the shipped final Day 9 declaration scaffold and day-local test runner honestly
• separate current deterministic evidence from the deferred cumulative capstone and executable Phase B
• delete the orphan control-recovery chapter that describes retired APIs
• align the Day 2 task groups, Day 5 counter ownership, and cumulative reviewer workflow

Scope

The final cumulative scope is exactly thirteen paths: ten navigated Week 4 chapter rewrites, deletion of book/src/week4-06-control-recovery.md, docs/week4-day-split.md, and the corrected Day 5 test-helper docstring in tests_refsol/test_week_4_day_5.py. Relative to the reviewed predecessor 9063a9ad, the exact successor changes only that one docstring line. All other executable, starter, runtime, reference, test, benchmark, fixture, evaluator, configuration, Weeks 1–3, preface, SUMMARY/navigation, publication, and release bytes remain unchanged.

Verification

• native builds and installation check
• 497 reference tests passed with 8 intentional skips
• all 146 Week 4 tests and the focused Day 5 6/6 checkpoint passed
• mdbook build, sitemap, link, diff, exact-scope, and preservation checks
• hosted macOS run 33317437557 succeeded on exact head 22e091ea31e30aa737837d22a5db6f6c86d1d74d
• fresh exact-head learner, factual/preservation, and correctness reviews all returned GO

The cumulative runner, capstone, evaluator, and validation-ownership repairs remain explicit Phase B work and are not claimed by this prose-first patch.

AI-Assisted: GPT-5.6 Sol + Sentinel

Reviewed-by: GPT-5.6 Terra + Scholar (learner)

Reviewed-by: GPT-5.6 Sol + Oracle (factual consistency)

Reviewed-by: GPT-5.6 Sol + Sage (correctness)