I backed up a bit and built a 2D model so we could perform some quick testing. This model is fairly unstable due to the explosive nature especially in the first few iterations.
With that in mind I decided to work my way up in solver setting complexity to determine what setup would be required. The only way I could get the model to be stable and give what looks to be a fairly reasonable solution required settings that does not support our Motion package. With that in mind, I would think you could study the silencer design fairly well without even including the bullet as part of the simulation. Trying to run this sort of model with motion would be an extermaly long runtime.
Check out the attached 2D model that you can run out on your end.