I don't fully understand spark gap. I have pieced together a few facts:
Sine wave TIG power in AC when viewed on an oscilloscope, is a curvy line. Each half cycle, current tapers down to zero, then gradually builds in the other direction.
When we weld aluminum, most older transformer welders are balanced flow current, or 50/50 balance. As argon is not relatively conductive, it must be ionized to easily conduct. Imagine a bolt of lightning. If we could see faster than we can, we would perceive steps in a bolt of lightning. A space in ionized air conducts, electrons stop at the edge of ionized air, billions of electrons rush to this point, but are unable to move until sufficient power of electrons ionize another volume of air. Then these electrons surge forward, and the process repeats. In a bolt of lightning, each step happens so fast that our eyes can't see each step. We see it as a crooked bolt of lightning.
Welding in AC, many transformer based welders are fixed at sixty HZ. direction of electron flow reverses 120 times per second.
Electrons flowing from the pointed tungsten ionize easily. Electrons flowing from the workpiece find it more difficult to ionize shielding gas. There is more surface they are spread over, and if welding aluminum, oxide layer is less conductive than clean aluminum.
If we didn't have HF, an overrunning high voltage, low energy high frequency power source, ionization of shielding gas would have to be recreated 120 times per second.
A 1960 car used collapsing magnetic field to induce high voltage, low energy electrical charge, igniting gasoline/air mixture. Something similar happens in the HF circuits in transformer welders. The spark gap fires, dissipating a build up of electrons, then pressure falls, interrupting flow. The magnetic field in a transformer coil collapses. This collapse, brings the magnetic field across transformer coils, inducing high voltage.
varying gap changes the intensity of this process. greater HF increases radio interference. There are things you can do to reduce a problem.
Spark electrodes should be polished. sandpaper makes thousands of scratches, interfering with ionization.