> As a physicist, the moment when everything just clicked was when I realised that connected Feynman diagrams were basically the cumulants of that distribution.

And the generating function of the cumulants is the logarithm of the generating function of the distribution (Fourier transform).