To review why these would be useful, let χ[T](s) = ⟨ exp(i ω T) ⟩, then χ[A + B] = χ[A] χ[B] for any A and B which are independent, so the characteristic function for a uniform random variable is going to be χ = (exp(i ω) − 1)/(i ω) and thus the sum of n independent uniform random numbers is just χⁿ. Meanwhile the PDF of a random variable is just the inverse Fourier transform of the characteristic function, so we have
fₙ(t) = ∫ dω/(2π) exp(-i ω t) [χ(ω)]ⁿ
Pr(n samples > 1) = ∫{1 → ∞} dt fₙ(t).
You could then interchange those two integrals, solving the convergence problem of exp(-i ω ∞) by giving ω an infinitesimal negative-imaginary component. This translates to moving the ω contour a little below the real line, and then the normal tricks of complex analysis should give you definite values for all of the integrals as residues about the simple pole at 0.
There are lots of relationships between the jump times of the Poisson process and subdivisions of the real line, so it's possible that the post is doing some version of a valid calculation, but without the right reasoning/understanding.
Using simple methods from martingale theory we can show that the expected value we're looking for must be no smaller than 2 and no larger than 4, but I'm in no mood to delve into more detailed calculations.