> Metric expansion does not result in causality issues because it's an isolating phenomenon
No, in the specific case of the concordance cosmology (and more general expanding cases such as an expanding Robertson-Walker spacetime, or a 4-dimensional de Sitter spacetime), the manifold is still Lorentzian [1] and globally hyperbolic [2].
These features let one sort trajectories into spacelike, null, and timelike. If spacelike trajectories are forbidden to interacting matter and gravitational waves, then we get our usual idea of causality [3].
Collapsing Robertson-Walker spacetimes and anti-de Sitter spacetimes are defined on globally hyperbolic Lorentzian manifolds. They are defined as free of matter and gravitational waves. However we can perturb them slighly by adding a bit of matter, even matter that generates small amounts of gravitational radiation, and as long as that matter and radiation moves only on timelike trajectories or lightlike trajectories, we have the normal causality even though the interactions among separated bits of normal matter perturbing these collapsing spacetimes will generally become more frequent and stronger over time. Indeed, this remains the case even for very highly accelerated compactions, where at late times distances between objects contract faster than light between the same objects can move.
Other commonly encountered spacetimes -- the flat spacetime of Minkowski approximately describes laboratories on Earth and in the International Space Station, and the (exterior) Schwarzschild spacetime which approximately describes the spacetime in which Earth-orbiting satellites move -- are also globally hyperbolic and Lorentzian, and thus we only need to keep superluminality at bay to guarantee normal causality.
In short: our usual idea of causality comes from being able to forbid a certain class of trajectories, and in physically realistic spacetimes (i.e., ones which approximately match parts of our observed universe), those are all as far as we can tell free from things moving spacelike (i.e., nothing's faster than a massless particle, and light is made up of massless particles, so nothing is superluminal).
Taken the other way around: our local observations of causal interactions (both locally, like in laboratories, and astronomically) are strong evidence in favour of us inhabiting a manifold equipped with the requisite features [1][2]. Alternative explanations have generally either failed in some way (self-inconsistency, obvious incompleteness, or inconsistency with repeatable observations) or been more complicated than, but reduces to, General Relativity in some limit. None of the second variety of alternatives has yet to be demanded by actual observations that cannot be explained using General Relativity as the underlying spacetime theory.
- --
[1] https://en.wikipedia.org/wiki/Pseudo-Riemannian_manifold#Lor...
[2] https://en.wikipedia.org/wiki/Globally_hyperbolic_manifold
[3] https://en.wikipedia.org/wiki/Causal_structure