For t statistics with microscopically small p values, the convert-to-z-score functions puts in z=13 -- which is a ridiculously large z-score; anything over z=6 has a p value smaller that 2*10^(-9). It never seemed worth the effort to accurately compute the precise z-score corresponding to a large t statistic, since both values are way out in the tails, and if you really believe data is that normally distributed, I have a black swan for sale.