I can see how the paths of the backward and forward waves can have different lengths, although I didn't realize that they must be different. What got me thinking, though, is that the virtual particle's Past emission/absorption node exists at a less entropic state than its Present emmission/absorption node. I sketched out a simple Feynman diagram with an arrow of time below it, and realized that in an expanding universe, the arrow of time also represented an entropy gradient. It seems to me that the entropic difference between the Past and Present nodes needs to be expressed somehow. After all, we expect to see some kind of emission whenever a particle drops to a less-energetic state. Why should we not expect something similar with respect to virtual particles when the nodes at which they create/annihilate are at different states?
Am I missing something really elementary?