[ajmurmann]

I think time and trust is so critical. I remember a talk by the people who made FoundationDB about their approach to testing. They spend a year just creating a variant of C++ that allowed them to control concurrent execution order and build a framework to simulate/mock network, disc access etc. It's incredible work and AFAIK more than paid for itself. Yet, at most places it would be impossible to get a year-long testing effort of that scale funded or really any engineering effort that produces nothing shippable. All too often you might get 3 months and after 2, a urgent product request comes up and the work sits and rots away till it's useless and nobody gets back to it