I believe that the exciton energies should respect the symmetries of the excitonci hamiltonian regardless of the number of k-points I include in the `nscf` calculation. I tested the BSE calculation and inspected whether for cubic systems such as LiF and cBN the exciton energies are degenerate according to the symmetry representation. LiF has 48 symmetry and is not problematic. I saw a small degeneracy breaking of the order of meV only when including GW. On the contrary cBN breaks the degeneracy when performing a BSE calculation. I performed various tests such as including or not `rim_cut` and `rim_W`. I tried as well to change BLongDir and LongDrXs or using flags such as `Em1Anys` and `OptDipAverage` without success. I include the input files here [cbn_share.zip](/uploads/0467cde64904c35770470d3cd85fbad7/cbn_share.zip) Clearly, there are some subtleties on how the dielectric function is computed in the q->0 limit. We know that \epsilon and \epsilon^-1 are not well defined in this limit, while W is well defined. I tested the code and I believe the issue arise on how \epsilon^-1 is computed in the code. In particular, it might be related to the wings (G=0, G'!=0) and (G!=0,G'=0) component of \epsilon^-1. Indeed for LiF those numbers will be real as G->-G' for inversion symmetry. While for cBN not. I'll perform further tests and inspect the code but perhaps, you already have some suggestions or feedbacks on the matter.