I asked VWD about this : why can't we change this parameter and the number of iteration once we start a simulation. I assume that what we want to do with VWD (perfecting one still render) was not the original purpose (making animation) so we are coming with new needs. What I do is systematically change both before clicking simulation: it can have an impact on the performance depending of your configuration

My experience on self collision is that we don't need to be perfect. If you see all my renders in this post there are some but you can't see them! For my current project with a two layered cloak I have plenty of self-collision but with the textures, the size of the cloak in my target render it doesn't matter.