Have made many of them, including practical 50,000V air core resonant capacitor charging power supply.
as to your questions, flash over spacing:
it depends on your primary voltage, air breaks down at 70,000V per inch (under controlled circumstances) but because homemade capacitors don’t control corona discharge this distance is much greater. I suggest that you experiment. Also the primary circuit can have much higher voltages then the primary transformer actually provides I would design the capacitor at least 4 times as resistant to the voltage then what your transformer will give you. That said, for 12,000VAC I usually go with 1" of space around the plate (not submerged in oil). Note that 12,000VAC has a peek to peek of around 16000V
As far as the plate stack, yes you can assume that it will be 40 mil, if you build a clamping device, if you just stack it, it will be taller. either way you will want to measure it as the calculations depend on it. You can clamp it with wood and 3/8" plastic all thread. or nylon with plastic all thread.
Most PE film I have used resists voltages of 1500V/mil minimum. It varies greatly. I might make a test jig. I assume that you are using a neon sign transformer. You should have an autotransformer supplying the power to the system so you can adjust it, If you don’t have one I suggest getting one. Anyway to setup the jig you simply get two blocks of dry wood, put some aluminum foil tape on the two blocks, put the sheet between the plates and connect the plates to the neon sign transformer. Use the autotransformer to bring the neon sign transformer online, turn the dial up slowly, and note when the dielectric breaks down. Do this several times and use the lowest value that you tested to.(unless it was a defect in the PE film).
I might use a 40 mil thickness to make a 12000VAC Polyethylene capacitor but that is just me not pushing the limit. at 10 layers, you should have enough redundancy to avoid failure due to small film defects.
Be safe.
