The parish stirs.
The six ferals now wander around their homes, notice nearby enemies and pursue them. Escape their territory and they walk back. Each still carries one finite soul.

The fourth and final step in Shaun’s chosen sequence gives the existing ferals small places to inhabit. The spell family, reacting scenery and creature attacks are already in the build; now approaching a quiet husk can start an encounter.
A small place to haunt
All six original ferals roam around their own home patches. Their patrol goals stay within two metres, with slow scuttling and small pauses. Actual movement drives the existing Blender gait and positioned footsteps. No new generated mesh or asset service was needed.
They notice visible, reachable hostile wizards or creatures within six metres, from either faction. A short creature cue marks acquisition. Their original homes and detection bounds leave both starting wizards outside immediate camp aggression.

Close enough to matter
Pursuit uses the actual terrain and navigation at three metres per second. A feral still prepares for three-tenths of a second and deals nine damage at contact. Getting out of range, finding cover or interrupting it with a hit can prevent that contact.
Following brood from either faction can defend against an aggressive camp. It leaves quiet camps alone, and existing courier, ritual and movement orders keep their priority.
It knows where home is
The chase stays within a twelve-metre home leash. Losing a valid target, solid-cover visibility or a usable route starts the return. A blocked creature retries its way home; it does not teleport. It finishes returning before acquiring another target.
Returning does not heal the feral. Killing one clears it for the match and releases its one soul beside the actual death position. There are no automatic neutral respawns or extra rewards: eighteen souls remain accounted for throughout the match.

The quiet clearing before
The protected 0.5.14 build ran the same camera and movement sequence. This front husk was previously passive and stationary, so approaching it did not trigger pursuit. The comparison uses the two saved Mac packs.

Ready for the combined playtest
The full regression passed 1,214 checks and verified twenty-three audio files. The exact pack passed 235 asset checks, 85 camp checks and 98 combat checks. The camp scenarios include both factions, cover, raised unreachable targets, blocked pursuit and return, pause, interruption, deaths, restart and conserved souls. A thirty-second scenario also ran the normal rival with camps enabled.
The new Mac build and an independent checkpoint are saved, with the complete previous delivery retained. Direct native app inspection remains unavailable while the Mac is locked; native rendered fixtures, exact-pack checks and standalone executable startup passed. Human combat feel and art/audio judgement, physical Intel, Windows and notarization remain unverified.
All four approved increments are now together. The next step is Shaun’s hands-on playtest: how encounters begin, whether the Brute warning reads clearly and how the completed spell family feels in a fight. Selectable gods and wizards, another creature roster, multiplayer and true terrain deformation remain future decisions.