> It's equally true that this class of mistakes represents an increasingly vanishing share of real world exploits.

Is this true? It seems implausible to me.