Q1: A typical Riemann surface has no holomorphic automorphisms, and this implies a negative answer to Q2. I don't see that Q3 can work: the p-subgroups surely don't uniquely determine the group in general.
The literature on these questions is quite large. It is a little hard to tell, but some of the material on the inverse Galois theory problem (which does use curves) might help you.