Release v0.1.9 - #1195

No actionable comments were generated in the recent review. 🎉

_{Comment @coderabbitai help to get the list of available commands.}

Adversarial review record (release machinery)

Verdict: the diff is exactly release machinery plus one gated oracle hardening; no stray changes.

• Surface: 11 files — the same 10-file surface as the v0.1.8 release commit (CHANGELOG, HISTORY, README, docs/index.html + the four generated site assets, pyproject.toml, uv.lock, vera/__init__.py) plus scripts/check_doc_counts.py.
• Version: check_version_sync.py reports "Version 0.1.9 is consistent across 6 files"; tests/test_release.py 47/47. The docs/index.html VeraBench caveat deliberately still says v0.1.8 — it records which version the v0.0.18 sweep ran against, a historical fact, not a version-sync surface.
• CHANGELOG: [Unreleased] cut to [0.1.9] - 2026-08-04 (16 fix bullets + 1 new Added bullet), fresh empty [Unreleased], both compare links updated (v0.1.9...HEAD, v0.1.8...v0.1.9).
• HISTORY: one v0.1.9 row in the Stage 19/20 table, single issue link per the row-format gate (which caught my first draft's 9 links and was obeyed, not weakened).
• Oracle hardening (test-first): FAQ.md's headline test count was provably blind — corrupting it to 9,999 passed the pre-change oracle — and the new check flips RED on the same corruption and GREEN on the true value. This closes the drift that shipped twice before.
• State cross-checks: open bug issues == {ch05_closure_nat_return trapped once in a full conformance run (unreproducible in ~960 attempts) — suspected rare runtime/GC flake in a CI gate #996} == the KNOWN_ISSUES Bugs table; limitations sync green (43/18/16/7/3); doc-counts green (8,840 / 136 / 176 / 42); no closed-issue rows left in ROADMAP; no strikethroughs anywhere.
• Workflow: on merge, release.yml sees the strictly-increasing version on main, validates the [0.1.9] section, builds, and pauses at the pypi environment for maintainer approval — no manual tag or upload is part of the path.

Per the standing instruction, this PR is converged and left for the maintainer to merge.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.79%. Comparing base (5053677) to head (e767803).

@@           Coverage Diff           @@
##             main    #1195   +/-   ##
=======================================
  Coverage   93.79%   93.79%           
=======================================
  Files          99       99           
  Lines       34134    34134           
  Branches      458      458           
=======================================
  Hits        32017    32017           
  Misses       2104     2104           
  Partials       13       13           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

Converged — ready for your merge

Final state at head 1e69d1ec:

• CI: all checks green (0 pending, 0 failed); merge state CLEAN.
• CodeRabbit: success at head. Two findings across two rounds, both fixed and answered inline: (1) the FAQ check now errors when the headline pattern is absent instead of silently skipping; (2) HISTORY.md's "By the numbers" release total was two releases stale — corrected to 205 and now cross-checked against README's status row by the oracle. Both fixes mutation-validated. The fresh pass at this head added nothing.
• Audit: 2 inline findings (0 open), 0 outside-diff findings, "Actionable comments posted" sum matches exactly.
• Version: 0.1.9 consistent across all 6 sync files; test_release.py 47/47; doc-counts, site-assets, and limitations-sync gates all green.

Per your instruction I have not merged this one. On your merge (merge commit, per the release convention), release.yml will detect the version increase, build, and pause at the pypi environment for your approval before tagging and publishing.

Re-converged at 9d5fa1a0 — ready for your merge

This head folds in the fixes for the four CodeRabbit findings that posted on #1194 after its merge (relayed by the maintainer): the CHANGELOG bullet's conformance count now reads "(176, was 175)"; a set-equality pin ties the test file's KEYWORDS tuple to _KEYWORD_FN_NAMES - _HOST_INVOKED_FN_NAMES (mutation-validated); and both E153 rationale branches now state precisely that no unqualified call site can reach the declaration and that the formerly half-usable module-qualified route is deliberately closed. The MD038 finding was declined with rationale (38-row nesting convention, no markdownlint gate). All four replied on #1194.

Audit at this head, taken after CodeRabbit's status reached its terminal state: CI 0 pending / 0 failed, merge state CLEAN, CodeRabbit success with its fresh pass adding nothing, ledger 2 inline findings / 0 open / 0 outside-diff, "Actionable" sum exact. Test total is now 8,841 (the new pin), propagated through TESTING/FAQ/README/ROADMAP with all oracles green.

Still yours to merge; the pypi gate waits on the other side.

@CodeRabbit full review