Research Tailcat as data-plane prior art

[forkwright]

Context

Tailscale has released Tailcat, a small data-plane-focused system that uses WireGuard, NAT traversal, and DERP without Tailscale's control plane. Its architecture may provide useful prior art for Hamma's gated data-plane work, even though Tailcat's product shape and trust model differ from Hamma's control-plane-compatible roadmap.

Official overview: https://tailscale.com/blog/tailcat

Research questions

• How does Tailcat bootstrap peers and encode/share connection state?
• How does it handle identity and key material, endpoint discovery, DERP selection/fallback, NAT traversal, reconnection, and persistence?
• Which architectural seams or protocol observations apply to dictyon and mitos, and which are specific to Tailcat's control-plane-free model?
• Does Tailcat expose useful independent-oracle or interoperability strategies for Hamma's blocked Phase A gates?
• Do any findings warrant changes to Hamma's public contracts, ADRs, or existing implementation issues?

Constraints

• Treat Tailcat as prior art, not as code to port.
• Preserve Hamma's clean-room Rust boundary: distinguish public protocol/behavioral facts from implementation details, and do not copy source.
• Keep data-plane activation blocked under the existing Phase A contract until its declared gates and completion authority are satisfied.

Deliverable

Produce a concise, evidence-linked research note that separates verified facts from inference, maps relevant findings to Hamma's architecture, and recommends either concrete follow-up issues or no change.