In your notation, "f_0", "f_3" and "f_5" are CONSERVED moments, thus their relaxation rate MUST be 1 -- collision does NOT affect the conserved moments and that is the manifestation of conservation laws in terms of microscopic dynamics.
The relaxation rate for "g_0" does not seem to be correct.
These are the conserved moments: \rho, ux, and uy, right. By definition, the equilibrium MUST be the function of conserved variables ONLY. Thus, the equilibria of conserved quantities are themselves.
Usually, there is an interface force in the multi-phase/component models (e.g., SC model), thus, the relaxation rate for the momentum affects the force and that is the defect in the LBGK model.
For models with reactions, the relaxation rate for \rho destroys the mass conservation.
The whole point of MRT is to set the relaxation rate correctly for each moment (or mode). This is also based on the basic linearized Boltzmann equation in kinetic theory (Gross-Jackson model) and projection method.