@whuber you asked a question about using FFTs for deriving poisson binomial PMFs in a comment. I ran into this not too long ago and your convolution approach was very helpful. You might find my conclusions interesting.