And it's worth mentioning that both approaches make sense for the application being considered. When you are sending a suite of purpose-built instruments and a

And it's worth mentioning that both approaches make sense for the application being considered. When you are sending a suite of purpose-built instruments and a one-of-a-kind rover to Mars, taking advantage of a narrow launch window, to accomplish a very well-defined scientific mission, you require the spacecraft to be safe. If you are sending a bunch of less-expensive spacecraft to perform a less intensive investigation of a bunch of asteroids, you might be more risk-tolerant. But as the cost of the instrument suite goes up, you'd have to be more risk averse. Additionally, if you expect to be in a difficult radiation environment, things change. LEO is not a difficult environment, but deep space is. It's a systems engineering problem, and there is no single correct approach. It's an engineering question, with the usual set of tradeoffs, expertise, and modeling. NASA and private industry have been studying this for decades now. (I spent a couple of years working on a software-implemented fault tolerance scheme for science algorithms in spacecraft.)