Caption: The horizon problem illustrated. The diagram is not-to-scale.
Features:
These starting points, in fact, form a sphere centered us called the last scattering surface (LSS).
In the diagram, the LSS is the big sphere.
where r_proper is proper distance, r_comoving is (cosmological) comoving distance, a(t) is the cosmic scale factor, and t is cosmic time measured from time zero at the big bang singularity of the Friedmann-equation Λ models---which is before by a fraction of second when the Friedmann-equation Λ models can be true.
Comoving distances are constant with respect to cosmic time.
For the usual convention that cosmic scale factor a(t=present) = 1, the proper distances are equal to the comoving distances at cosmic present = to the age of the observable universe = 13.797(23) Gyr (Planck 2018).
So the radius of the cosmic present observable universe is always ∼ 14.3 Gpc ≅ 46.6 Gly in comoving distance.
At that time cosmic time, the cosmic scale factor a(t)/a_0 = 1/(1+z) ≅ 1/1100.
So the observable universe was much smaller then. The expansion of the universe caused it to grow to its size at cosmic present = to the age of the observable universe = 13.797(23) Gyr (Planck 2018).
However, its comoving volume was the same as now.
The fastest a signal could travel relative to its local inertial frame (which would usually be at least approximately comoving frame) is the vacuum light speed c = 2.99792458*10**8 m/s (exact by definition) ≅ 3*10**8 m/s = 3*10**5 km/s ≅ 1 ft/ns.
In fact, the farthest back in cosmic time, we can even imagine running the Friedmann-equation Λ models) is ∼ Planck time t_plank = sqrt(ħG/c**5) = 5.39125*10**(-44) s when they would have a density ∼ Planck density ρ_Planck = c**5/(ħ*G**2) = 5.15500*10**96 kg/m**3 (see Wikipedia: Planck units).
The LSS was in self thermodynamic equilibrium (very nearly).
How could this be if those points are NOT causally connected?
There is NO way that the points could be in near thermodynamic equilibrium (i.e., at nearly the temperature) by influencing each other in the Friedmann-equation Λ models.
Other evidence shows the observable universe at recombination era t = 377,770(3200) Jyr = 1.192*10**13 s (z = 1089.80(21)) was very homogeneous (and isotropic) in all respects.
However, cosmologists and physicists in general do NOT like the idea of a beginning of time out of nothing (i.e., ex nihilo)---especially with fine-tuned initial conditions.
So the extreme homogeneity of the early universe is a problem which is called horizon problem.
The name is NOT so good since it has to be explained. The horizon of a point is sphere at the limit of causal connection and somehow points beyond the horizon have extreme homogeneity in the observable universe which is has to be explained. A better name than the horizon problem is the homogeneity problem, but that name is NOT much used.
For an explication of inflation paradigm, see Cosmology file: inflation_paradigm.html.
Of course, the inflation paradigm creates new problems---but physicists like new problems to solve.