I am simulating a model several times (say 100) using simul_replic in stoch_simul. I would like the simulated series of some exogenous shocks to be identical across the 100 simulations. In principal the paths should be identical in terms of some random shocks, and not for others.

I guess there could be a way around this using set_dynare_seed. Then I would like to ask whether once can fix the seed for only some shocks in a model with multiple shocks.

Suppose we have this in a model with two shocks:

for i=1:200
set_dynare_seed(i);
stoch_simul(periods=200, noprint, nograph);
g_sim(:,i) = g;
pi_sim(:,i) = pi;
end

How can we have the path of shock 1 be the same for all 200 paths, but different random paths for shock 2?

That is not possible. The seed affects all shocks. You would have to program this yourself using the simult_- function that underlies the stochastic simulations.