This issue can be nearly-completely resolved. Namely, as in Stein-Weiss' book on Fourier analysis on Euclidean spaces, there is a sharp bound on the ratio of sup-norm to $L^2$-norm of spherical harmonics of a given degree. In fact, the result admits considerable abstraction...! That ratio is (up to a constant depending on measure, which obviously affects $L^2$ norm but not sup-norm) something like the square root of the dimension of the corresponding irreducible repn of the orthogonal group.
(This replaces the all-too-happy estimate $|e^{i\xi x}|\le 1$ in the tiniest case.)
Some simple algebra (!) shows that the dimensions grow polynomially (in an explicit way...).
So, yes, ignoring details to give a qualitative answer, there is indeed a very explicit good Sobolev theory on spheres. My notes https://www-users.cse.umn.edu/~garrett/m/mfms/notes_2013-14/09_spheres.pdf do lots of background, and write out the case of sufficiently high Sobolev index implies continuity, as an example.
Addition after OP's comment/question: here, as in other cases where we have a Laplacian eigenfunction expansion, with sufficient info about the eigenfunctions and growth rate of their eigenvalues... an "expression" $f=\sum_n \langle f,u_n\rangle\cdot u_n$ for a "function" $f$ in terms of an orthonormal basis $\{u_n\}$ for $L^2$ consisting of eigenfunctions for the Laplacian immediately gives all the $H^k$ norms, even for general $k\in\mathbb R$.
Namely, first, for the definition $|f|^2_{H^k}=\langle (1-\Delta)^kf,f\rangle$ with $0\le k\in\mathbb Z$,
$$
|f|^2_{H^k} \;=\; \sum_n |\langle f,u_n\rangle|^2\cdot \langle (1-\Delta)^k u_n,u_n\rangle
\;=\;
\sum_n |\langle f,u_n\rangle|^2\cdot (1-\lambda_n)^k
$$
where I've changed notation so that $\Delta u_n=\lambda_n$.
Second, in fact the right-hand side of the latter display makes perfect sense for $k$ an arbitrary real number.