Duro Kurepa conjectured that the function on the title is always nonzero in $\mathbb{Z}/{n \mathbb{Z}}$ provided $n>2.$ Daniel Barsky and B\'enali Benzahgou [MR2145571 (2006a:11025)] proved this. Thus, for all odd prime numbers $p$ $$ ku(p) = 0!+1!+ \cdots + (p-1)! \pmod{p} $$ is an inversible element of $\mathbb{Z}/{p\mathbb{Z}}.$
Question: There are infinite subsets $S$ of the set of odd composite numbers such that for each $s \in S$ one has that $ku(s) \pmod{s}$ is inversible in $\mathbb{Z}/{s \mathbb{Z}}.$ ?