Sounds like your enemy AI is stuck in a tight loop. First, check the loop condition – make sure there's a clear exit or a state change that happens after a few iterations. Add a counter or a timer so if it loops too many times it automatically resets or switches to another state. Also, look at the pathfinding logic; if the target node is blocked, the AI might keep trying the same spot. Adding a fallback or random waypoint can help it break out. And don't forget to log the position each loop – that will show if it's stuck on the same point. Once you see the pattern, you can tweak the condition or add a fail‑safe.

Sounds solid. One more tweak you could try is a “stuck detector” that measures the distance the enemy has moved over a short window—if that distance is below a threshold, you trigger the same wander transition. It’s a bit more granular than just a frame count and can catch cases where the AI is moving slowly but still stuck. Also, consider giving the AI a small “panic” mode that runs for a couple of frames and forces it to pick a completely random target before resuming normal behavior. That can break cycles when the pathfinder gets trapped in a corner. Anything else you’d like to dig into?

Sounds good. If you want to keep things even more robust, add a small “cool‑down” after a panic reset so the AI doesn’t immediately re‑enter the same trap. And keep an eye on the navmesh size: a larger sampling radius can help skip tight corners but might miss tight passages, so you could tweak that based on the level’s scale. Anything else on your mind?