[omniglottal]

Nothing is untestable. If validation results in destruction, then destructive testing is how this is done.

Failing to test is a choice, similar to the choice made by a software engineer who determines, without testing, that their production service has no bugs.

If a component's flaw can only be detected right before a failure, we need statistical analysis of multiple such failures being intentionally induced to estimate the component's limits.

Not finding or knowing about existing limits or flaws is the direct result of not looking for them.

[salawat]

THIS.

As a Quality Assurance person, this is what drives me up the wall, across the ceiling, and down the other wall.

You can always do a test run. Not doing that test run is essentially saying every person you subject to that system is not worth the trouble of doing the test.

[berniedurfee]

FOMO is a helluva drug.

[jjk166]

No one was claiming that it was untestable, the company did decide to go with an acoustic monitoring, which is a destructive test, instead of a scan, which is a nondestructive test. The company claimed there was no non-destructive test that could be done. The issue is that when you're dealing with short run custom components, like the hull of a submarine, just because one passes a destructive test doesn't mean the others are free of manufacturing defects. You can't do useful statistical analysis when your sample size is small.