From this site and this post, doubly even codes have transversal $S$.
Based on this post, surface codes don't have transversal $S$ gates. We can check the boundary stabilizers of surface codes are not doubly even, so surface codes are generally not doubly even and don't have transversal $S$.
However, what if we go to the toric codes which seems to be doubly even? What's the logical effect of transversal $S$ gates?