Google Research found that coordinated multi-agent teams outperform solo agents by 80%. Uncoordinated teams? Only 57%. That 23% gap is the coordination tax and it's measurable.

Mnemom's trust data makes it sharper: "Five high-performing agents with poor coordination can create more risk than a cohesive mid-tier group." The risk isn't weak agents. It's strong agents that don't work together.

A registry knows who's on the team. It has no idea how the team behaves. Group coherence requires runtime proof: who did what, in what order, with what outcome.

This is why proof-based identity isn't a nice-to-have. It's the missing coordination layer.