mccall_risk.md was never delivered — edition is 143/145 against source

[mmcky]

lectures/mccall_risk.md ("Job Search V: Modeling Career Choice" / risk-sensitive preferences) exists upstream in QuantEcon/lecture-python.myst but has never been delivered to this edition. It is absent from lectures/ and from lectures/_toc.yml here, and present in both upstream.

A full set comparison as of 2026-08-19:

	Count
lectures/*.md in lecture-python.myst	145
lectures/*.md here	143
Missing	learning_approximation.md (tracked in #210), mccall_risk.md (this issue)

How it went missing

It was delivered by #242, which was auto-closed on 2026-08-07 by the rebase automation's force-push race — head_ref_force_pushed and closed both at 01:34:27Z, with the automation's own "♻️ Automatically rebased after #230 was merged" comment nine seconds later, and no separate manual close in the event log. The mechanism is filed as QuantEcon/action-translation#280.

The gap then went unnoticed for twelve days. The closed PR's other five files were updates to lectures that already existed and have since been superseded by later syncs, so nothing about the tracker state looked wrong, and no stage in the pipeline compares the source lecture set against this edition's — not CI, not review mode, not the sync run. It surfaced only because the branch outlived its PR and a post-drain tidy-up went looking for stray refs.

Recommended recovery

A targeted gap-fill — translate init -f for mccall_risk alone, at the current engine — rather than recreating #242 from its branch. That branch (translation-sync-2026-08-05T23-15-41-pr-760, still present, 15 ahead / 37 behind) carries a 2026-08-05 snapshot of six lectures; merging or regenerating from it risks reverting twelve days of newer edits to the five that did land, which is the failure mode in QuantEcon/action-translation#276. A fresh gap-fill also gets a translation from v0.26.0 rather than the engine of the day.

Two things to check when it lands: that _toc.yml gains the entry in the right part, and that the lecture's figure captions and any {ref} targets are localised, since the caption classes in QuantEcon/action-translation#254 and QuantEcon/action-translation#255 are still open.

Once the lecture is in, the orphaned branch can be deleted.

[mmcky]

Update 2026-08-20 — this is now the edition's only coverage gap, and one premise in the body above is wrong

Coverage is 144/145, not 143/145. learning_approximation.md was seeded on 2026-08-20 in #261 (closing #210 and restoring its four local {doc} references), so mccall_risk is the single remaining gap and this issue is the single remaining tracker for it.

Correction to the body. It states that "no stage in the pipeline compares the source lecture set against this edition's". That is false, and the error propagated into a work plan before it was caught. Three mechanisms already perform or record the comparison:

1. The dashboard collector (collector/collect.py in QuantEcon/status-translations) runs a genuine set diff nightly on cron 17 3 * * *. It named mccall_risk in data/history/2026-08-06.json — the morning after the lecture landed upstream, a day before the force-push race closed its delivery PR — and in all 14 snapshots since.
2. The engine files a translation-sync-failure issue on the source repo per failed run. Fifteen were open estate-wide, all now closed.
3. The CLI's translate status classifies a source file with no target as SOURCE_ONLY, though action.yml exposes only sync/review/rebase, so it is unreachable from the Action.

The gap is surfacing, not detection — nobody reads any of them. That work is scoped on QuantEcon/action-translation#222.

Recovery route, updated. The body's plan — a targeted gap-fill rather than reviving #242's branch — still holds, and the case against the branch is stronger: it is now 40 commits behind main (15 ahead), so a revival would trip QuantEcon/action-translation#276's stale-snapshot class hard.

Per the 2026-08-20 ruling on that issue, the route is translate init -f mccall_risk.md against a source clone at origin/main — init reads source off the local filesystem, so it cannot carry a stale snapshot. Four practical notes from running the same recovery on four other lectures that day:

• Structural parity is flaky on this path. Expect refusals — one lecture took six attempts for two passes, always failing by dropping the leading ```{raw} header block. The guard is fail-closed and writes nothing, so retry, but verify the run you keep.
• init will revert all 19 localised _toc.yml part captions to English and add ToC entries for lectures this edition does not have. Revert its ToC entirely and hand-add only - file: mccall_risk.
• Caption and figure-label localisation is unreliable, not absent — one of four lectures came through with every caption and axis label byte-identical to English. Diff target values against source before opening the PR.
• The body's warning about a genuine CellExecutionError still stands: pull _build/jupyter/reports/mccall_risk.err.log from the failing run promptly, the artifact expires.