I won't burn resource on pure RNG events like that as what you said can happen, it's a meh event.

However, where is your data. And did you collect data in collaboration with others or did you burn 50k coupon/ingot on it yourself?

Do note that you need to avoid self-selection bias in your data collection. Intend, participation and methodology must be established before the actual collection process.

But that only says something about the output, it says nothing about the input. How do you know the input haven't dropped (because people figured out that it's bad and/or because the big ptw who would go for it already have the ninja)

That's just not an acceptable testing methodology/data collection.