On the connection between large deviations and the problem of reliable stabilization