Fun tool, but let's give it a go using maths.
As for solution I also ended up using Box-Muller like some other people in this thread (@kat and @danilang I both saw used it).
Still unsure what the "correct" choice of multiplaction value is when scaling the results, but a simulation at 10m only had 6 test outside the band, so felt that was an ok margin of error. A proper implementation would have to test for this.
Here you go! Unrelated but in the same category, are there any good guides for using the optimization tool? I have seen a few examples however I am really struggling to get my head around how to configure this correctly.