nx_comparestale_dependency_t138.nx

nx_comparestale.nx -- IS EVERY PUBLISHED COMPARE PAGE STILL TRUE TO ITS SOURCE? WHY THIS EXISTS. /compare/doctor sat a MONTH stale, publishing "THE root blocker: emitted ELFs carry no line mapping" long after that stopped being true -- with a perfectly valid generated_unix. Nothing could see it, because nothing compared the page to the thing it was generated from. A TIMESTAMP TELLS YOU WHEN A PAGE WAS MADE, NEVER WHETHER IT IS CURRENT. HOW. Both generators now stamp "source_hash" = djb2-48 over the .matrix/.sota they read. This organ recomputes that hash from the source on disk and compares it to the published page. Deliberately CLOCK-FREE: mtimes lie across copies (this tree has TWO knowledge/compare directories and I lost an edit to exactly that) and clocks lie across machines. Content does not. VERDICTS, and the distinction is the point: FRESH page hash == source hash STALE page exists and disagrees with its source -> the source of truth moved, publish did not UNPUBLISHED source exists, no page -> never generated, or gate-refused NOSTAMP page predates the provenance stamp -> republish once and it becomes decidable Exit 1 if any STALE. UNPUBLISHED does NOT fail the gate: a domain whose matrix gate legitimately refuses (legal, present>=5) must not be reported as drift -- CONFLATING "REFUSED" WITH "DRIFTED" IS THE SAME DEFECT THIS ESTATE KEEPS PAYING FOR. license_tier: ORIGINAL expect_exit: 0