A useful benchmark: For an event with probability 1/N, the probability this event will occur at least once in N trials is approximately 0.632 (converges to 1-1/e as N goes to infinity).
This may be the only useful expertise I can contribute to this discussion, so I will retire to the peanut gallery. Some related math in this thread.