Two lectures both read 工作搜寻 V — the five myst#760 renumberings were never delivered; two refactor backports also outstanding

The 2026-08-20 full-corpus audit (project-translation#50) content-verified this edition's drift. With mccall_risk now seeded (#264), what remains is small and precise — but the seed has turned the missed renumberings into an active collision:

The Job Search numbering (reader-facing)

myst#760 inserted Job Search V and shifted five siblings. The shifts were in auto-closed #242 (the action-translation#280 race) and never arrived by any other route. Now that #264 has correctly delivered mccall_risk as 工作搜寻 V, the edition has two adjacent lectures both numbered V (mccall_risk and mccall_persist_trans, _toc.yml rows 81–82), and everything after is one behind:

Five H1 lines (plus each file's translation.title and, for jv, the embedded {index} target). Note mccall_persist_trans's state file claims the change is synced — its 2026-08-18 UPDATE wrote a source-sha past the skipped hunk (filed as action-translation#287) — so this will not surface as OUTDATED in translate status.

Two small refactor backports, same lost window

• ak_aiyagari — myst#706 (+9/−9): hh → household inside the jitted helper bodies; target still has 23 hh.j_grid-style lines. The audit's reviewer also noted the current form silently binds the module-global hh inside functions that take a household parameter — inert today, wrong if reused.
• harrison_kreps — myst#669 (+7/−4): price_single_beliefs refactor (imbq_inv → solve-based A), typo fixes, bold→italic; imbq_inv still present ×2.

And one state-file-only touch

likelihood_bayes is the mirror case: content already carries myst#977 (simulate(a, b, rng, …) present, prange gone) but .translate/state points at the pre-#977 SHA. Left alone, a future forward run may needlessly rewrite a current translation — refresh the state file only.

All verified against source 285a1c6 / target b136ab3; per-lecture evidence in the audit's drift table.

Start with the audit drift table and the five named lecture files, then inspect _toc.yml rows 81–82 and each .translate/state entry. Check translate status before changing anything, and use the cited source and target revisions as comparison points. Done means the five numbering/title updates, both named refactor backports, and only the likelihood_bayes state refresh are present and verified.