Intricacies of Optimize Cross Section in Karamba

We're trying to use the "optimize cross section" in karamba to design a lightweight grid matrix structure based on member utilization (as opposed to displacement).  There are about 850 elements in the model and a list of 40 wood trapezoidal (rectangular) cross sections.  Is there a rule of thumb for how many iterations to use in the "opt cros sec" component?  The component will often return no errors for 20 iterations, but then will complain about 3 members needing to be bigger at 100 iterations.  Again at 300 iterations, no errors...

Also, the choice of initial cross-section to use when assembling the model seems to effect these error messages (larger initial section seems to work better).  

We'd like the analysis to include buckling at the member level and the model seems to be accounting for that satisfactorily.  Though, the Karamba manual does mention that "during the optimization of cross sections normal forces NII are not updated" and to "use 'Analysis ThII' to determine NII iteratively.  Is it indeed necessary to use the "Analysis ThII" component along with the "opt cros sec" component, and if so what is the typical setup for this?  

Any thoughts are much appreciated -