Ralph Loop 2026-03-16: Eventful Move Coverage

Change

Added replay-input diagnostics for realized move coverage:

1. research.prepare_replay.summarize_replay_rows() now reports: - moved_rows - large_move_rows - large_move_eventful_rows - large_move_untagged_rows
2. research.status and research.discord_status now surface the move-coverage summary for both stored loop state and the current persisted replay rows.
3. research/README.md documents the new diagnostics so Ralph can distinguish scanner misses from sleeve misses.

Why

PRD 001 is about eventful observation quality. Before this change, replay summaries told us how many rows were tagged eventful, but not how many rows actually moved later without being tagged. That made it hard to tell whether the current “no trade” state was a sleeve problem or a collection-quality problem.

Validation

Command:

PYTHONPATH=. pytest -q tests/test_prepare_replay.py tests/test_research_status.py

Result:

6 passed in 0.03s

Current bottleneck observed

From the current persisted replay snapshot:

1. total_rows=1280
2. eventful_rows=1
3. moved_rows=28
4. large_move_rows=28
5. large_move_untagged_rows=28

Interpretation:

All currently observed large replay moves are being missed by the eventful tagging layer. The next bounded PRD-001 improvement should target scanner/eventfulness calibration, not broader sleeve expansion.