I think this may be related to this post I found: Report on non-zero intermolecular Coulomb-14/LJ-14 from Gromacs/2023-gpuvolta and gromacs/2023
In short, I think my version (or perhaps just the my particular installation) has a bug in reporting intermolecular 14 interactions, where it reports nonzero intermolecular interaction energies for 14-LJ and 14-Coul. On my same HPC, I have access to a 2021 version of GROMACS and I can repeat my workflow on the same topology/gro/mdp files without this issue in the 14 energy components, so I don’t think its a problem with the simulation setup.