Agents walk back and forth + performance issues

Hi, finally updated the project + now using the latest beta-version of this project.
But in the performance-documentation:
Optimization - A* Pathfinding Project (arongranberg.com)

You mention to either remove the Update or FixedUpdate depending if the agents have a Rigidbody or not. And then remove one or the other.
Is it still true for the beta-version? Because, I can’t reach them.

And have reference to my assembly:
AStarPathfindingProject
PackageTools
ALine

  • saw the RVONavmesh component was deprecated in this beta-version, and said:
    “This component is deprecated. Local avoidance colliders never worked particularly well and the performance was poor. Modify the graphs instead so that pathfinding takes obstacles into account.”

So I removed it, but should i keep a RVO simulator + a Graph (use recast)
And agents with:
AI (using RichAI)
Seeker
RVO Controller?

Because now i think i see a slight performance gain with beta. But agents start to go zic-zac after they destroy a structure that had a NavmeshCut on it.(If i manually place them closer on their blue-lined path, they can snap out of it, with luck, but can’t do that in runtime :P).
So is my setup bad in this beta-version.
Or should I keep on trying different variable-values? Change AI to maybe the simple AIPath instead of RichAI?
Also tried have large EndReachDistance with no change.

And also read another thread, but not sure how to implement their solutions, e.g.:
AI Moving back and forth - Beginner Questions - Support Forum (arongranberg.com)
Can’t apply Vantalks solution as I’m using Recast Graph
And burdock solution 1 was my try above, solution 2 i’m not sure of.
And thought it would be a common issue that the package should have an inbuild solution by tweaking some numbers (I assume they are overshooting)

Oh and lastly, when they start to go crazy like that, they walk outside the navmesh as it wasn’t there (using RichAI).