I think the "right" way to calculate advective tendencies, if we don't want to rely on having either the non-physics tendency terms or the all-physics tendency terms as "special" diagnostics (which most runs will not save due to space constraints), is to actually use the output winds with neighboring profiles of T, q with the upwind differencing scheme to derive them.
The https://github.com/NCAR/ccpp-scm/pull/372 PR is being merged with the calculation of tendencies from UFS output already including the physics tendency.
I think the "right" way to calculate advective tendencies, if we don't want to rely on having either the non-physics tendency terms or the all-physics tendency terms as "special" diagnostics (which most runs will not save due to space constraints), is to actually use the output winds with neighboring profiles of T, q with the upwind differencing scheme to derive them.