One thing you might try doing to identify the source of the problem is to post process the last converged substep. If you include OUTRES,ALL,ALL command before solving, the results associated with all substeps will be saved to the results file. You should be able to post process these in POST1. If you look at the deformed configuration (e.g., PLDISP command) of the last converged load step, you might see what part of the mesh is on the verge of becoming entangled. I see you are already using SOLID285 (switching to that element type sometimes helps in situations like this, but you're already using it). Have you tried using a different criterion for NLADAPTIVE remeshing? In the example I sent you in your other post, I used an energy criterion which seemed to work pretty well.
You are navigating away from the AIS Discovery experience