This is almost certainly "board noise" - pretty common on YMB. And not even clear whether it was intended to be single or double sided. Try searching up the board for the ultimate source and its reliability.
It came from a poster I consider pretty reliable, and was purported to have been read from a copy of the protocol. The precise sentence was "An efficacy boundary that allows early stopping consistent with a one-sided level 0.001 O-Brien-Fleming design."
What matters is HR (and the associated p value) - relative medians are just a convenient proxy for them. And the ratio of 30 to 13 is much more aligned with the HR of about 0.55.
I'm sure you know what an HR of .55 at the 2nd interim in START would mean, and how it would be received in the press and by investors. It would rock the world. I'm not expecting .55, by the way, as I'm sure you're not.