Frozen after the initial42 rows, before the broader fixed-depth check.
The initial record-slope rowi2 cases have small Vmax/mu, so test whether that
was a favorable phase choice. This is an explicit adversarial follow-up,
not an independent confirmation or a pre-initial-data prediction.

At depth10,height2,actual record slope, test all completed-clipping boundaries
i1..6 and all three vertical copies d0..2. For every nonempty positive
marked component and the full box, use common windows floor(sqrt(Q)),
Q//64+1 andQ//8+1. Cap Q<=2^19 and one process120seconds.
Record all rows including means below10. The candidate sufficient bound
Vmax<=8mu is tested only as a deliberately loose finite conjecture on rows
with mu>=10; any violation is retained with the full path witness.
Do not infer a moment failure from a failure of this stronger sufficient
condition. Do not generalize a pass to other depths or phases.
