Hello again ronv and Flyback,
Well you are both right to a point...
When the contacts first open, the arc can occur as soon as they are just barely open. Once the wave goes through zero, the arc might stop. But by the time the wave goes high in the other direction the contacts may have only moved 1um or less, so the arc can start again. This can happen many times and depends on the speed at which the contacts open. If they bounce (and they will at least a little) then there is another mechanism for starting up the arc again. But even after the bouncing is complete the contacts may still be close enough together to allow another arc to form. It's all about the voltage breakdown of air over a distance. The smaller the distance the less voltage it takes to create a current flow, so they dont have to be touching for this to happen repeatedly. It's only once the arc is broken AND the contacts have moved far enough apart that the arcing will stop completely.
It's also interesting that the model for an arc in air is a resistance that starts out relatively high then decreases as the arc is formed. So the resistance between two partially open contacts may read high, but once they arc over the resistance goes low even over the same distance between contacts.
The material movement depends a lot on the level of current and this current is relatively low, but then again we dont want the small contact surface areas to start to pit because that makes the closed resistance higher and higher.
This kind of action can really be a problem sometimes with relays that have to switch something between two sources. If the arc lasts long enough one source can actually power the other source even though the contacts are not wired for that kind of operation. It's nuts when this happens because things blow up (like 120vac inverters).