It only goes up to n = 300, so I adjusted downward by 10% since n = 331.
The likelihood of being a control (flipping tails with a loaded coin ) is p = .33. Number of trials (flips of the loaded coin) = 300 (limit of the tool).
Range of successful flips/trials coming up for control arm is between x1 = 0 and x2 = 90 (the 10% adjustment because of tool limitation).
The likelihood of that result is .148 or roughly 15%.
The CRO would have been charged with doing a randomization within a gnat's whicker of 2:1. And we've ended up with just about 70:30. Thereby significantly weakening the statistical power. Cue start of lawsuits..
I just don't buy it, I'm afraid. There is another explanation.