I commented on this the other day; there's a useful NPR article on the subject.
> The night before the launch, Ebeling and four other engineers at NASA contractor Morton Thiokol had tried to stop the launch. Their managers and NASA overruled them.
> That night, he told his wife, Darlene, "It's going to blow up."
https://news.ycombinator.com/item?id=33488205
https://www.npr.org/sections/thetwo-way/2016/01/28/464744781...
He didn’t feel right about it, but that’s not the same as hard data. They didn’t actually have data on the performance of the material at those temperatures, which is part of the big discussion with his managers. He essentially said, “we don’t know how it will act in those conditions.”
Granted, on a safety critical system, having no data should be a scenario very, very carefully considered as a harbinger of doubt, but having no data isn’t the same as having a concrete understanding of the risk. The problem with large, complex systems is you will always find someone who thinks management isn’t taking a certain risk seriously. I know aerospace engineers who don’t want to fly because they feel they know the risk intimately, yet that’s a 5 sigma industry.
Columbia was similar. They had lots of data points that the foam shedding wasn’t a problem is different scenarios, but they didn’t have the same evidence about that specific scenario (the deltaV, flight profile, and where it struck).
If the threshold for risk/doubt is zero, they would never launch. It’s hard to build an engineering consensus without data
In the article, risk was known. "It would be career limiting." The managers knew the risk was to themselves because they weren't offering up a solution.
The Challenger was different. There wasn't hard data that characterized the risk. That was the main problem; there was an absence of known risk because the operational parameters were outside the testing scope. It was more like "We don't know what the risk is, if there is one." That's a very different animal because you're dealing with uncertainty and low probability events. Those risks tend to show up in hindsight. (same with Columbia for that matter).