> In psychology there is a definite boundary between individual behavioral dynamics and population behavioral dynamics.
> either you describe the individual or you describe a group
In gravitational physics, with respect to a flow in a dynamical system (an example is galaxies in an expanding universe) we can use a Lagrangian observer (e.g., one galaxy, drifting along with the flow, tracing out a pathline/worldtube that depends on features like its mass-evolution and proper motion within a cluster of galaxies) or a Eulerian observer (e.g. a notional observer with no spatial motion at all, watching alllll the galaxy clusters jiggle, swirl, turn, and age a little differently in relation to her). One can convert observations of each type of observer to the other in a rigorous mathematical procedure, since they are just two (families of) the infinity of different observers allowed by even just Special Relativity. See e.g. <https://en.wikipedia.org/wiki/Lagrangian_and_Eulerian_specif...> for more detail.
>> There are boundaries between where GR and QM are predictive
> this feels alien to me
You can do both quantum matter and classical General Relativity in one of several effective field theories, which I'll return to below.
GR and relativistic quantum field theories (QFT) purport to make accurate predictions in strong gravity, which one only finds deep within black holes (i.e., not on our side of any horizon), but they make very different predictions in that regime pretty generically. Generically in the sense that choosing different behaviours of particle-particle interactions (and self-interactions) do not really move the needle on GR's prediction of a collapse to a core of infinite density. However, in various approaches which convert GR's classical gravitational waves into large number of gravitons, one can write down a matter QFT in which charged particles' self-interaction can lead to a degeneracy pressure (a repulsive force) that increases at higher particle energies such that they overwhelm gravitational collapse at very high but finite density in black holes of arbitrary mass.
In weak gravity, like we have in our solar system, QFTs allow us to prepare significant masses in superpositions of (spatial) position. General relativity does not allow for such superpositions. We are approaching lab-testability, with results from sensitive accelerometers allowed to point at tiny superposed masses.
However, in regimes far from (non-negligibly) gravitating superpositons and strong gravity, GR and QFT are usefully (and possibly fully) compatible. We get good results in astrophysics from semi-clasical gravity, where the classical curved spacetime of General Relativity couples with the expectation value of QFT matter (i.e., we average out some quantum weirdness and justify this by the weak gravitational effects of the "lumpy" weirdness being practically impossible to measure; superpositions and ultra-high-energy/ultra-high-denisty systems might be too lumpy).
We also get good results from perturbative quantum gravity and canonical quantum gravity, for example. Neither of these latter two is really classical General Relativity so they can deal with the gravitation of superposed matter (otherwise they give for all practical purposes the same answers as semiclassical gravity). These approaches do not work in strong gravity, however. Essentially they become calculationally intractable or they crash into unresolved problems splitting spacetime into space and time (in order to do time-dependent quantum mechanics).