Wouldn't a better experiment be to roll one 10K times and see if the sides have equal probability? Who cares if the dimensions are a little off if the effect on the roll is minimal, or conversely are even the best toleranced dice still biased?
Check out the "skew die" https://www.mathartfun.com/DiceLabDice.html
And a video https://www.youtube.com/watch?v=uAnCL3vhVIs
It would be more impressive to see a die with a pattern of a buckyball with the pentagons and hexagons having equal probability of being rolled. I'm sure its theoretically possible, just not practical.
They're not regular - the faces are not regular polygons - but that's not a requirement to have an equal chance of landing on each side. The standard d10 is an example of this in itself.
The skew dice faces are all identical but none are even parallel.
"Gamescience dice are more consistent than the X-Wing dice", with some analysis regarding the flashing from the mold suggesting that sanding it smooth will increase the consistency.
https://www.awesomedice.com/blogs/news/d20-dice-randomness-t...
They _manually_ tested Chessex and Gamescience dice by rolling them thousands of time. They found both brands' dice to be off. Gamescience dice had a big dip on the number 14 face, which is the face that has the bit of flash where the die is cut off the sprue by hand.
I suspect it's just easier to make a fair d8 than a fair d20.
That probably goes for other die sizes also. I was looking for a d14, d16 and d18 and while I found the first two from Gamescience, I could find a d18 only from one brand, other than Gamescience, so I'm guessing that Gamescience decided they simply couldn't make a fair d18 work. I mean, why skip d18 when you have d14, d16 and d20?