I will first address the string theory part of the question.
String theory provides examples of physical systems admitting several descriptions that provide natural bridges between Kleinian singularities (and therefore Platonic solids), ALE spaces, quiver diagrams, ADE diagrams and two dimensional Conformal Field Theories.
The scene is given by compactifications of string theory on Kleinian orbifolds $M_\Gamma=\mathbb{C}^2/\Gamma$ where $\Gamma$ is a discrete subgroup of $SU(2)$. The space $M_\Gamma$ admits a Kleinian singularity at the origin. After studying this physical system, one is less surprised to see that Kleinian singularities, quiver diagrams, ALE spaces, ADE diagrams and 2 dimensional Conformal Field Theories all admit the same ADE classifications since they provide different descriptions of the same underlying physical system.
Michael Douglas and Gregory Moore have studied the compactification of string theory on Kleinian orbifold $M_\Gamma$ using D-branes as probes of the geometry. D-branes are extended objects on which strings can end. D-branes provide a physical description of the geometry in terms of supersymmetric gauge theories. Such supersymmetric gauge theories are efficiently summarized by a quiver diagram with a very natural physical interpretation: the nodes correspond to D-branes with specific gauge groups on them and the links between the nodes are open strings ending on the branes.
The minimal energy configurations (the vacua) of these supersymmetric gauge theories are obtained finding the extrema of a potential whose construction is equivalent to the hyperkhäler quotient construction of Asymptotic Locally Euclidian Spaces (ALE spaces) first obtained by Kronheimer. ALE spaces are HyperKähler four dimensional real manifolds whose anti-self-dual metrics are asymptotic to a Kleinian orbifold $M_\Gamma=\mathbb{C}^4/ \Gamma$. Physically ALE spaces described gravitational instantons. ALE spaces provide small resolutions of the Kleinian singularities where the singular point is replaced by a system of spheres whose intersection matrix is equivalent to the Cartan matrix of an ADE Dynkin diagram. One can also consider Yang-Mills instantons on such spaces. The gauge group associated with the Yang-Mills instantons is given by the type of ADE diagram obtained by the resolution of the singularity. This was analyzed in the math literature by Kronheimer and Nakajima. Physically the ALE instantons moduli space is equivalent to the vacua of the gauge theory description of D-branes located at the singularities.
The link between D-branes on ALE spaces (or equivalently Kleinian singularities) and the ADE classification of two dimensional Conformal Field Theories (CFT) was studied by Lershe, Lutken and Schweigert. Although the geometry is singular, the CFT description is smooth. The 2 dimensional CFT is coming directly from the string description: as a string evolves it described a 2 dimensional surface called the string worldsheet. D-branes enter the CFT as boundary states. In the description of the CFT, one recovers Arnold's ADE list of simple isolated singularities.
Updates
I would like to comment on the non-stringy part of the question. This is motivated by the comments of Victor Protsak.
If one remove all the string theory interpretation in the discussion above. What is left is Kronheimer's description of ALE spaces. Kronheimer's construction provides a beautiful realization of McKay's correspondence between Kleinian singularities, their crepant resolutions and the ADE diagram. This is reviewed in chapter 7 of Dominic Joyce's book "Compact Manifolds with Special Holonomy". From that perspective, the string theory description provide a physical interpretation of Kronheimer's construction and adds a natural link with quiver diagram and 2 dimensional Conformal Field Theory.