What makes you believe that it is not necessary to converge with respect to k-point sampling and mesh density? To me, there is no particular difference between SE and DFT in that regard, except that mesh density can be set lower for SE SCF calculations than for DFT SCF calculations.
