For the first question - yes. For a simple model - no.
The biggest issue is that it takes several (4-6) hours for it to solve to that point, and I can't just analize this one specific step, because the aim of my simulation is to predict welding distortion of simple butt-welded plates. The simulation crashed half way through the calculation process (i.e. the weld was 50% complete). What bothers me the most is that I don't get a chance to look at what failed (like residuals or unconverged solution deformation) to try and fix it. Thus I am having an "educated" guess - changing contact stiffness factor, simplifying material properties (turning off bilinear isotropic hardening, messing with coefficient of thermal expansion), trying to constrain the model from moving as much, turning stabilisation on, trying adding some contact step control, making the mesh dense, simplifying so that only the added weld material has nonlinear properties etc... If I mess with the simulation too much it gives me some results e.g. 17mm (0,7 inch) deflection instead of expected 4-6...
It is a core part of my master's dissertation and the time is running out... I don't know what else I could do to make it work.