Wouldn't you have to mathematically account for a 2 to 1 randomization to reach assuredness of being at the median before unblindnig?
Ignore the medians. Look at some cardiac trials for statins - they get perfectly valid HRs and p values without either arm getting anywhere near median survival.
Medians are a distraction - and lead to all sorts of misunderstandings. It is the number of events and the HR that matters.
Market Data 
Markets 
AI
