A common special case is the approximate three-body problem in which:
In this case the orbit of body B0 is a conic section (ellipse, parabola or hyperbola) with the barycenter of the body B1/B2 system as the conic focus, and one can express the orbit using Keplerian elements relative to this barycenter.
If the small terms are not neglected, they can be treated as perturbations so that the barycentric Keplerian elements are slowly varying functions of time.
In the limit m2/m1 --> 0 the problem reduces to the pure Keplerian one where the focus is the center of body 1 and we can say (if e<1) that body 2 is `in orbit around' body 1.
If m2/m1 << 1 then we can still use body-1-centered Keplerian elements, but they will vary with time more rapidly than the barycentered ones.
For space probes in orbit around the Sun, there's a difficult choice to make here. The effect of Jupiter is enough to make the solar system barycenter (SSB) noticeably move relative to the center of the Sun. The mathematically natural choice is to express the orbital elements barycentrically. However, one cannot see the SSB - the astronavigationally natural choice is to ask `where am I relative to the Sun', even though those elements are time-variable. The same is true of the Earth-Moon system - choosing geocentric or Earth-Moon-barycenter (EMB) coordinates.
In general I will choose heliocentric and geocentric coordinates rather than barycentered ones.
See also the discussion of the Hill Sphere in ems.html