Many advancements in science have happened because we stopped for a second, and then looked to generalize our assumptions. Consider,

e.g.

Euclidean geometry -> non-euclidean geometry; Classical analysis -> nonstandard analysis; Linearity -> non-linearity; Homogeneity -> inhomogeneity; Flat spacetime -> curved spacetime; Singular probabilities -> superposition.

All of these were loosening of certain criteria that opened up many possibilities. It is certainly erroneous to assume we must, by necessity, have a homogeneous cosmology.