This effectively claims that administrators are biased into making Martingale style doubling bets. As an administrator of a large program, cancelation may mean

This effectively claims that administrators are biased into making Martingale style doubling bets. As an administrator of a large program, cancelation may mean the end of your career "you're the guy who wasted X billion dollars on Y". Meanwhile many ineffective courses of action will eventually achieve their goals sufficiently well given enough resources. As a successful administrator you are likely to double down on failures rather than let them fail. With the goal to recoup your losses from the first round. This tends to mirror my experience with large organizations. Projects tied to a manager are rarely canceled unless the manager leaves (willingly or not). I wouldn't be surprised if unicorns had a similar trajectory of justifying previous losses with even bigger bets.