nx_vsroll_gate.nx

nx_vsroll_gate.nx -- proves a domain lifts on its WEAKEST leg and refuses on any unmeasured one. The teeth that matter are T4 and T7. T4 is the anti-Goodhart tooth: the roll must take the MIN, so a domain with one EXCEED axis and one PRODUCTION axis earns PRODUCTION. T7 is the tooth that makes the roster worth having at all: ADDING a scope must be able to LOWER a domain and never raise it. If declaring more of what a domain means could improve its grade, the roster would be a lever instead of a ruler, and every domain would be described in whatever terms flattered it. T5 is the cross-domain leak: a claim on scope `latency` under `search` must not satisfy scope `latency` under `network`. Sharing an axis NAME is not sharing the WORK. license_tier: ORIGINAL No hw writes (Rule 26).