simplify_sum ('sum(1/binomial(n, k), k, 0, n));
sum(1/binomial(10, k), k, 0, 10);
Thanks to Ed Scheinerman (sage-support newsgroup) for pointing it out.
Maxima uses the Zeilberger algorithm to attempt to find the closed form. Zeilberger returns something which looks OK but simplify_sum misinterpreters the result.
I have added a stronger check for zeilberger sums in commit [a04d8e]. Maxima now returns the sum unevaluated.