The Truss Diagonals were critical for fatigue.  Above is a summary of the computed stress ranges.  Configruation 00 was the original and C12 was the final. Material thicknesses were adjusted between these, and the final configuration actually ended up lighter than the original through observing the interactions and sensitivities of the weld stress to both the diagonal and node can thickness' and balancing those for the best result while keeping the number of material thickness' to a minimum for the structure.  

All of the determination of the transverse stresses and summary across the load cases to determine the node with the maximum stress hot spot stress range was performed in FEWeld

Depiction of local coordinate systems calculated at each weld joint node in FEWeld so that stress tensors are automatically aligned with the transverse, longitudinal, and out of plane directions of the weld joint

The nodes were more critical for the extreme load cases - more distributed and greater plasticity, still with a healthy margin against the limit state.  The upper nodes were used for the validation of the FEA model against the API requirements because the lower nodes had the floorbeam connection which, first, introduced most of the load that caused the higher stresses, and second, is not a classical tubular joint and difficult to analyze classically.  The FEA model predicted that it has a margin of 2 against the limit state given the results of the validation of the upper node.