How much accurate is description of nuclear fission rate by means of Kramer's formula?

We compare results of dynamical modeling of the fission process with predictions of the Kramers formulas. For the case of large dissipation these are two: the integral rate R_I and its approximation R_O. As the ratio of the fission barrier height B_f to the temperature T reaches 4, any analytical rate is expected to agree to the dynamical quasistationary value R_D within 2 %. We performed modeling using four different potentials and found that the difference between the R_O and the R_D sometimes exceeds 20 %. Such discrepancy is not acceptable nowadays because it is comparable to the quantum, non-markovian and multidimensional effects. The features of the potentials used which cause this disagreement are identified and studied. It is demonstrated that this is the R_I, not the R_O, which meets the expectation above irrespectively of the shape of potential.