workflow-fix: verify_plan check — paired-contrast rows need per-context sources on both arms

Overview / Motivation

Auto-filed by the workflow-fix-on-bug protocol from prose follow-ups surfaced by BOTH methodology critics + the codex statistics critic + the consistency-checker reviewing #810 plan v13 (four independent catches of one gap).

Goal

Add a verify_plan.py check that every registered paired-contrast row names a per-context data source on BOTH arms.

Workflow gap

• Bug observed: #810 plan v13 registered a 9-row paired bootstrap whose named full-side pack (uh_summaries.pt) lacked 2 of the 9 rows (im_end/turn_nl); the registered success criterion was unsatisfiable from the named inputs and the falsification leg was uncomputable exactly when those rows win. Caught only by reviewer legwork (pack row-name reads).
• Why it is a workflow gap: verify_plan has no statistical-input-existence check for paired contrasts; the reuse-fitness (c) required-cells idea exists for artifacts but not for per-row paired-DV sources.
• Confidence (emitter): medium-high (four independent reviewer catches)

Proposed change (candidate diff sketch — refine in planning)

• In scripts/verify_plan.py: when a plan registers a paired contrast (grep paired + a row list),
• require a declared per-context source per arm covering every registered row
• (mechanizable shape per the critics: set(registered_pair_rows) ⊆ union(keys(full_side_sources))),
• FAIL when a registered row has no declared source on either arm.

Scope / surfaces

• Primary target: scripts/verify_plan.py
• Sibling: tests/test_verify_plan.py

Constraints / invariants

• Workflow-surface only; workflow_lint + ruff pass.

Provenance

• workflow_fix_target: scripts/verify_plan.py
• fingerprint: fe9772faa82a

Surfaced prose (codex-critic methodology, #810 v13, verbatim): "This should become a workflow verifier check for paired contrasts: every registered row must have per-context sources on both arms."