By quite a bit, right? Assuming passenger malfunctions were statistically independent, I guess passenger malfunctions would follow the binomial distribution.
Say there are 100 passengers. We roll the dice 100 times, once for each passenger.
The success rate for the overall flight would be 0.999 ^ 100 = 90.4% -> About a 10% chance of flight failure.
If there are 400 passengers, we get 67.0%, about a 33% chance of flight failure.
Given enough passengers per flight, few or no flights reach their destination.