Not going to open the file, I am too busy simming my own work and it would take me too long to debug it even if I wasnt busy but here are some quick questions
Do you have self intersections on any bone/muscle at any frame? You must deal with this if you do
What is your solver?
What is your integrator?
I find these solver / integrator pairs don't equal the same result
Have you tried increasing substeps? What is your current substep?
What is your collision point spacing? Have you considered lowering this?
What is your max newton iterations? Have you considered increasing this?
Have you considered putting selective sliding attachments on that indented area?