Problem with fatigue analysis and MAT ADD EROSION
I believe your implementation is fine as long as your units are correct. But I'd like to question whether removing elements is the correct approach here. It seems like putting a band-aid on a bullet wound.
Most of my Dyna experience is in dynamic impacts, with some implicit here and there. I've never looked at fatigue analysis, nor anything to do with drilling. But I'll try to help regardless of my lack in specific knowledge.
Have you tried changing the material model of the drill to something more simple? Say *Mat_Elastic with some basic parameters that somewhat represent the material, just so you can see if the problem persists? If it does, then you know it's not the material model.
At that point, I'd look into the mesh. Have you tried reducing the mesh size or changing the ELFORM of the tetrahedrals? Maybe try ELFORM 13 if you are using 10?
If that doesn't work, then look into the contact perhaps? Maybe there's something funny happening as the drill passes through the plate.
These are my thoughts.