Restivo's conjecture is false, but the correct bound is still not known. Recently Gusev and Pribavkina have produced a set of counterexamples for which the bound is 5k2 - 17k + 13 for k ≥ 4.
A related open problem is the computational complexity of determining, given a finite set S of words over Σ, if Fact(S*) = Σ*. Here the complexity is measured not in terms of the total size of S, but rather, the length of the longest word in S.
References:
V. V. Gusev and E. V. Pribavkina, On non-complete sets and Restivo's conjecture, in G. Mauri and A. Leporati, eds., Developments in Language Theory, 15th International conference, DLT 2011, Lect. Notes in Comp. Sci., Vol. 6795, Springer, 2011, pp. 239-250.
-- JeffreyShallit - 20 Jul 2011