You'd expect longer numbers in larger bases. The estimate at Wikipedia (https://en.wikipedia.org/wiki/Polydivisible_number#How_many_...) can be generalized: let F_k(n) be the number of n-digit polydivisible numbers in base k. Then F_k(n) ~ (k-1) * k^(n-1) / n!. This gets bigger as n increases up to n = k, then it gets smaller. If I'm doing the asymptotics right you have F_k(ek) approximately equal to 1 - so in base k the largest polydivisible number should have about ek digits.

The length of the longest polydivisible number in base k is in the OEIS (http://oeis.org/A109783) along with this conjecture.