Only testing the extremes didn't work

This was interesting from Ars:

https://arstechnica.com/uncategorized/2024/12/nasa-says-orions-heat-shield-is-good-to-go-for-artemis-ii-but-does-it-matter/

The thing that struck me was that the skip approach to landing the Orion capsule was what caused the problem of the big cavities in the Orion heat shield.  This was supposed to minimize the highest temperatures (and also g-force on the astronauts) but that ironically led to the problem by causing the internal heating of the shield.  To me this means that while the shield was tested on the ground for extremes, it wasn't tested as NASA decided to use the shield.  As I recall the original Apollo re-entry did NOT use the skip method so that might be why this wasn't seen before.

This reminds me a little of IC testing.  We always did the tests at the databook limits but never at levels either below or above the limits.  One thing that we could do on the Sentry IC tester was employ what was called Schmoo plots (named after the Li'l Abner character).  This would tell you where the part would pass/fail.  I remember one case where a card designer had used a chip beyond the data sheet limits.  The shop people told me that one vendor worked but the other didn't.  Actually what I found was that one vendor failed more gracefully (the plot was linear) while the other failed like a dog leg, linear but with an abrupt change in slope above the limit.  The shop guys weren't happy but I didn't ding the "dog leg" vendor just because their part didn't work ABOVE the limit.

However maybe I should also have been looking for holes in the plot.

More on Schmoo plots:

https://en.wikipedia.org/wiki/Shmoo_plot

More on the Sentry IC tester:

https://mtsi.substack.com/p/testing-the-z80-chip-with-a-1970s

Best Regards,
Chuck, WB9KZY
http://wb9kzy.com/ham.htm