You have a lot of nodes for such a small object. You might want to decrease the refinement at some points as you have over 8 million nodes now as the coarsest mesh which is still a little overkill. Ideally you’re looking at say a million nodes but you can afford extra nodes due to the computing power available.
Do note you should have a y+ of 30 to 200 for a wall function approach while for a full resolution y+ should ideally be less than 1 but LES is computationally very expensive so you might want to just use a wall function and stick to the less expensive y+ of 30 to 200. I see that you’ve done a proper BL inflation for Mesh 3 so lets see how that goes.
If you want to significantly decrease the number of nodes and increase simulation time I suggest doing what I’ve done here for my compressible project and using Tet mesh with an extruded plate containing a hole that represents your shape instead. Although note that I’ve been running into strange problems with compressible flow but it doesn’t hurt to try.