There are a few things you seem confused about.
First, if you have a arc that generates a pocket of oil-vapor that is submerged in a pool of liquid oil, it can never ignite, as there is no oxygen there.
With regard to quenching, basically the idea is to "stretch" the path of the arc until it breaks, and is as such quenched. There are actually high-power circuit breakers that specifically force a jet of air through the arc path when they open to quench the arc as rapidly as possible.
Basically, the liquid is supposed to boil, as this provides a source of gas. Also, it's worth remembering that the phase-change from liquid to gas requires a lot of energy, and this helps absorb the heat generated from the arc.
It is true that if a oil-immersed breaker cycles often, it will use up most of the oil, both through vaporization and carbonization, but the device should have sufficient oil to last for it's rated cycle-life.
There are a couple of things going on here.
First of all, many isolation transformers do NOT isolate the ground terminal on the load side; they simply connect it through to the ground on the supply side. I've never understood this reasoning, but there it is. This means that the input ground of your scope is still connected to the building ground even when using the transformer, which sort of misses the point.
The second thing is that unless the isolation transformer is built with an internal electrostatic shield, there is capacitive coupling between the primary and secondary windings, which can easily pass the 29 µA of current that you're seeing. This is why if you measure the voltage between either terminal of the secondary and the building ground, you'll often see a reading of about half the line voltage.
Best Answer
Nynas brand transformer oil is pretty common in this voltage range is rated at 25kV/mm.
When further "processed" in-house it can be up to 75kV/mm.
Typically what happens is when transformer factories get a truckload, they spill out the 1st bucket's worth then fill up their tanks and then sample test it with a slow AC ramp tested with predefined shape smooth electrodes and gap and then measure the breakdown voltage and repeat 10 times.
The results are often all over the map from 25kV to 45kV. Test houses like Heismann will usually void the 1st 10 tests and then take 10 more.
The reason for the variability is due to contaminants like a few particles of dust. Now imagine normal clean office air has 100k particles per cubic foot. Now figure out how you can avoid air and dust.
That may be impossible, even to have a perfectly clean container. Tote boxes are the best idea, not your own.
Chances are your gaps are around 10mm so you might think no sweat. Wrong. What happens is the ionic "dust particles" even in AC are attracted to the electrodes which freely release electrons under an electric field and any ionic "dust" or moisture particles may see a different localized charge level due to a difference in conduction and dielectric constant. Air is 1, Oil is 2 and H2O is 80.
What happens best case? Nothing. Typically? A slow relaxation oscillator in hours, minutes, or seconds detonates the charged particle as it slams into the electrode and goes "tick." Worse condition: it ticks at a steady rate of once per second or once per cycle (like corona) but still may be a quiet tick sound but starts generating H2 in the oil which will try to rise and release. Normally preventive or "Condition Based Monitoring" of HV assets means annual sample testing of oil for dissolved combustible gases like H2, H2C2, H2C4, H4C8. The higher the detonation energy, the higher the activation potential and release potion of the gas.
If you are lucky or good, nothing will happen. But this ticking is called Partial Discharge: PD is the Canary in the mine shaft so if you are aware, no worries. Any AM radio off channel will pick up the tick noise like lighting.
Many are now using processed vegetable oil as it can absorb higher humidity. Bigger transformers than 10MVA usually have air separator bladder bags to keep air out and release Hydrogen.
For a clean solution, use extremely clean high grade isopropyl alcohol that dries on glass without any residue. bake in oven at 180 deg for a few hours Then test with a slow HV ramp for PDIV threshold (Partial Discharge Inception Voltage). Then if normal quickly apply silicone coatings or high quality silicone sprays and build up a layer in a dust-free environment (class 10k or better, like a HEPA laminar air flow booth.)
But for giggles try anything and test for PD activity to get experience and become an expert in this field.
PD rise times vary from picoseconds to nanoseconds, so the RF signature goes past UHF.
P.S. I investigated one local transformer factory that had supplied many to wind farms in the < 10MVA range from 20k to 50kV and they were failing DGA sample tests for H2 levels rising above the lower explosive limit (LEL). It was a million dollar liability issue, so they invited me to find out where the problems were.