Add falsifier_one_cycle_resolvable axis 0-3 to meta filter_score rubric

Extend FILTER_SYSTEM (line 19-47) to require an additional field 'falsifier_one_cycle_resolvable' integer 0-3 in the judge's JSON output: 3 = resolvable in dry-run against historical rows today (no waiting); 2 = resolvable within 1 engine cycle; 1 = needs 2-4 cycles to accumulate data; 0 = needs ≥5 cycles or external real-world signal that AE doesn't already collect. In scoreProposal (line 104-121), after parseJsonLoose, validate the field is integer-clamped 0-3. Persist into the meta_filter_score move output JSON. Add a soft rule: if falsifier_one_cycle_resolvable === 0, append axis_concern { axis: 'falsifier_speed', concern: 'falsifier requires ≥5 cycles', severity: 'major' } but do NOT auto-DROP. The field becomes Commander-visible signal in the report. Keep binary KEEP/DROP verdict unchanged.

Re-score the last 19 historical filter_score moves with the new prompt: expect 3-6 to land at 0-1 (e.g. 2026-06-13's calibration proposal admits 'target ~3-4 cycles' — would score 1; 2026-06-12's S151 RECENT_ADMITS proposal needed '30 admissions ~3-6 weeks' — would score 0). Distribution surfaces a previously invisible axis-coverage gap.

Re-score last 19 historical filter_score moves with new prompt. If 0 score below 2, the LLM judge cannot distinguish falsifier latency — axis collapsed; drop. If ≥15 score 3 (max), the rubric is too lenient; tighten the 3-tier definitions or kill. Target band: 3-6 below median across 19 rows.

Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.