A widespread misconception in Statistical Mechanics is using microstates (microscopic states) instead of equilibrium macrostates (states of thermodynamic equilibrium).
A system of a few particles may show fluctuations corresponding to microstates typical of low-entropy macrostates. For example, a finite temperature cluster of 32 atoms may arrive, during its dynamical evolution, at a configuration close to a perfect fcc structure. However, this would be a temporary fluctuation between many other configurations. Thermodynamic entropy is an equilibrium property, i.e., a property of the ensemble of microstates; it is not a property of a single microstate. Therefore, it is improper saying that the cluster at the instantaneous fcc configuration has lowered its entropy.
Of course, at very low temperatures, the same system will remain close forever to the fcc structure, provided it is a local minimum of the potential energy.
Similar considerations apply to the question if there is a finite probability a thermodynamic system of a few particles could get the absolute zero of temperature as a temporary fluctuation. Temperature is an equilibrium property, thus depending on the whole set of microstates associated with the equilibrium macrostate. We can follow the evolution of the kinetic energy of a small system which shows fluctuations. However, a local fluctuation cannot be considered representative of the equilibrium average. Therefore, a fluctuation reducing the kinetic energy to its minimum value cannot be considered a fluctuation of the thermodynamic temperature of the system toward zero.