It could exhibit partial discharge, but then that discharge moves the voltage across the dielectric, and no further action results. In other words, the tiny capacitance of the air gap (~0.05pF?) discharges, rather than the full load current. It's like the static crackle when rubbing a balloon on hair.
If this happens enough times to cause both solder mask layers to break down, then a proper arc can start. Since the solder mask thickness is highly variable, this could be many kV more than intended!
Not that spark gaps are at all well controlled to begin with, but that's just making bad to worse.

It's worse still when you consider the second strike: the solder mask has been pierced, so now it's only an air gap. Except a small amount of solder mask has been carbonized, and more on each strike after that. Rapid tracking ensues, and you get a glowing burning ~short circuit.

Tim