np.random itself with that configuration can go negative with a probability of 0.62% as can be derived with scipy.stats.norm(loc = 7500, scale = 3000).cdf(0) or by looping a million times and counting the negative numbers.

Maybe the f2d function filters negative values but it sounds like a simple float-to-double conversion.

I'm not sure whether it was intentional, but, contrary to the headline, this random value was used to update the fund size daily as shown in the rest of the code. So, a single day for which the fund actually decreases wouldn't matter much. It might even be beneficial to make it look more real.

The function doesn't calculate the insurance fund amount directly, it only applies a change to the insurance fund (line 28), based on last day's trading volume. Presumably this function gets run every day, so the size of the insurance fund would be the sum of this process repeated multiple times. The chances of that going to negative would be astronomically low.

Since it's a normal distribution with a mean of 7500, and a standard deviation of 3000 multiplied by daily volume—it would only cause a negative change less than 1% of the time. This "PublicInsuranceFund" truly embodies their culture of "number go up".

This is obviously just speculation, but there probably wasn't a whole lot of thought put into it by the coder. The goal was "produce a fluctuating number based on daily trading volume so that it looks like we have a real insurance fund". The randomness is simply there to help make the fraudulent representation less obvious. The fact that the number can (very rarely) go negative is just another example of the fly-by-the-seat-of-your-pants thinking pervasive at FTX.

I'd imagine that the person writing it thought the function generates numbers 7500 +/- 3000 and not anything outside this range.