Restructure the V2 article around its causal explanation

[jamiepratt]

Orchestration

Issue kind: child implementation
Parent: #16
Depends on: #18
Type: AFK
Parallel group: article-narrative
Conflict risk: high
Expected touchpoints: V2 executable article source, math/explanation helpers, interactive assets, article-specific styles and data

What to Build

Restructure the V2 article into a standalone 2,500–3,200-word causal narrative for an educated lay reader with good mathematics but no specialist statistical or psychometric background. Lead with the paradox that frequency is a better structural idea while the complete scorer remains insufficiently reliable. Preserve the article's explanatory and executable richness behind progressive disclosure.

The always-visible core has six sections: the paradox; why V2 should help; how a better model can fail; the failure microscope; what the diagnostics establish; and the decision. The full series list becomes a compact disclosure.

Acceptance Criteria

• Always-visible narrative is 2,500–3,200 words and readable independently of earlier series articles.
• Logistic curves, priors, coverage, and MAE each receive a compact visual or prose refresher rather than a full inline derivation.
• A model-versus-complete-scorer pipeline prevents the failed gate from being misreported as evidence that frequency contains no signal.
• A curve-area/summed-probabilities explanation connects the response curve to the vocabulary-total estimand.
• The failure microscope distinguishes illustrative learners from repeated-run evidence.
• Compact supported-cell and coverage-versus-length views reflect the frozen diagnostic artifacts.
• Corpus construction, derivations, priors/grid details, selection, simulation protocol, historical gate, stress evidence, provenance, and code remain accessible as technical depth.
• Established findings, supported mechanisms, counterfactuals, and unresolved causes are labelled distinctly.
• V2 remains non-promoted and V1 remains the current target.

Validation

Run executable assertions and configured CLJ/CLJS validation; verify diagnostic artifact identity and source links; perform a focused render; audit visible word count and heading hierarchy; inspect every disclosure and interactive state in a separate browser tab.

Notes

Do not quietly change V2 parameters, its gate, or historical evidence to improve the story. The article may explain new post-hoc diagnostics only with explicit provenance and limits.