Completion-driven reviewer orchestration

Always continue background reviews only from substantive completion notifications; never poll, duplicate, overlap, or edit an active reviewer’s scope. Retry a first empty result once unchanged with an evidence-or-blocker contract; after a second empty result, replace only that reviewer with smaller disjoint scopes. Preserve completed evidence, retain exactly one pending follow-up while work remains active, then add fail-first regressions before production fixes, run targeted/full verification, and obtain fresh independent correctness and security reviews on the exact final tree. Chose completion-driven orchestration over proactive checking because polling, overlap, or patch drift invalidates review evidence.