Should the pressure and temperature coupling in Gromacs (tau_p and tau_t, respectively) reflect the size of the integration time step? 

For example, in a CG simulation you may have the .mdp arrangement:

dt = 0.03 

tau_t = 2

tau_p = 12

Whilst, in the atomistic arrangement:

dt = 0.002

tau_t = 0.5

tau_p = 0.5

What are the underlying risks/concerns of having a pressure or temperature coupling very small when using a large time step, and vice versa?