[pjerra]

Round 2 on 19z19: the Zul'Farrak run wiped twice at step 26 (the camp at 1666.77 903.85). Cause (PLAN-ZF-STEP26): the changed route pulled the camp before the pair [5648,7247] standing between the party and the camp; the camp step had no approach row, so the run leader stopped 30 yd short (15.4 yd from the pair's 5648) and the pair aggroed during the walk. Fix: antusul order = Witch Doctor 1633.3 868.3, the pair (patrol block kept), the camp, the north pack, as in the 19z16 clear; approach points 1620 880 z 8.97 (pair) and 1630 878 z 8.93 (camp), navmesh floors. New check internal/dungeon/approach_clear_test.go: a pull's stand (last approach point, or 30 yd short of the group centre from the previous group) must stay 20 yd from later groups (23 computed, 3 yd drift) and 15 yd from patrol paths; it FAILS on the old order (mutation proof) and fails only for the two guarded Zul'Farrak groups, other findings (other plans, 31 more in Zul'Farrak) are report-only and not fixed. Reviews: Opus cold review (SHIP). KNOWN LIMITS (from the review, not fixed here): (1) the 155626 patrol still passes the camp: with DungeonRunPullStand on, the camp pull's firing stand lands about 1.6 yd from the patrol path, so the 19.6 yd figure holds only for the approach point; cause 2 of the step 26 wipes (the patrol) needs module support for patrols; (2) the north pack's stand is 9.2 yd from the patrol path (report-only); (3) creatures in no group are not checked. Zerillis (10082) stays in the skip lists (agent-led driver only; the exporter drops skip lists, so no module row changes).