Yeah, stress concentrations at the weld toes are pretty much expected, and they can definitely depend on the mesh. Tweaking the mesh might help, but if the high stress is more about the weld geometry itself, it might not make much of a difference.

I’d go with sub-modeling for the welds—it lets you refine the mesh in critical spots while keeping the rest coarser, so you don’t burn through unnecessary computation time. Plus, it improves accuracy. Also, don’t forget to double-check your boundary conditions and load cases since they can seriously affect the stress results.

Weld toe stresses are one of those tricky things in FEA, so if you can, real-world testing is always a good call. Strain gauges can give you some extra confidence in the design. And if this thing is going off-road, definitely keep fatigue in mind—those conditions can be brutal!