"Systems drift when long-term consequences remain abstract.
The most expensive failures are quiet ones: a series of decisions that were defensible in the moment and disastrous in combination. In a scaling issue where efficiency and resilience pull apart, nobody wakes up wanting the incident; they wake up wanting the shortcut.
In a scaling issue where efficiency and resilience pull apart, everyone will focus on intent. The system only feels behavior. You see it in a cache added for speed that quietly changes correctness semantics. A clean stop line prevents optimism from keeping a failing bet alive long after the evidence turns.
Albert O. Hirschman is a reminder to audit the argument, not the intention. Write the failure story forward while the decision is still reversible, then insist on one guardrail. Keep it short enough to remember under stress.