It's very tempting to casually say that the discharge is the return path, but it would be more correct to say that it is a path of resolution which dilutes an imbalance of charge created by some other means.
We tend to think of current flow circuits but actually, circuits are just one particular behavior of charge.
Another is the accumulation of charge (say, triboelectric, ie, feet on carpet charge). This is a "static" charge - absent a pathway to neutralization, the excess or shortage of electrons "just sits there" on a charged object.
When the charged object comes close enough to an object connected to a large reservoir of more neutral charge (either via wiring back to the earth itself, or simply a conductive object having its own charge capacity such as a doorknob) then the electric field resulting from the difference in charge may in exceed the breakdown voltage of air for a specific geometry, and a spark may result.
In the case of diodes intended to present a shunt path for ESD, that reservoir of more neutral charge could anything from an actual grounding system, to the bulk of the devices's own supply rails/network. Ultimately an ESD event is a brief current spike to equalize charge, and since the possibility of damage from over-voltage is ultimately about over-voltage relative to other semiconductor segments equalizing charge with the powered or unpowered supply rails is sufficient.