There is an absolute maximum for the time step size, but no formula for directly calculating what minimum is the correct choice. The default choice of OptiFDTD is the aforementioned maximum. This at least guarantees a stable simulation.
I think your issue might be the number of time steps and not the individual time step size. Resonators and cavities trap light within them and it takes time for the light to leak out (Q-factor). This means that you will need to check to see if the energy in the resonator has had time to leave. You can do this by using an observation point placed inside the resonator or at on of the ports. You should see the electric field decaying with something similar to an exponential envelope. Check to see if the electric field has diminished to an appropriately small value by the end of the simulation.