The Ronin example perfectly proves your human-factor thesis.
Lazarus didn't break 256-bit cryptography; they broke a hiring process. A fake PDF bypassed every technical safeguard in the bridge architecture because the vulnerability was never in the smart contract code. That distinction matters more than most security discussions acknowledge.
Your point about burn-and-mint shifting risk to messaging layers is spot on but often overlooked. Removing the honeypot doesn't remove attack surface; it just moves it from vault custody to proof verification. ZK-bridges reduce trust assumptions mathematically but still depend on honest key management during deployment ceremonies. The human layer persists even when the cryptographic guarantees improve.
Atomic swaps being "impenetrable but impractical" is the uncomfortable truth interoperability keeps avoiding. Security and UX remain inversely correlated, and no amount of protocol innovation has fully solved that trade-off yet.
The $2B lesson isn't that bridges are fundamentally broken. It's that we kept treating operational security as secondary to architectural elegance.
RE: ENG/ESP: The $2B Hacks: The Hidden Anatomy of Cross-Chain Bridges | Los Hacks de $2.000 Millones: La Anatomía Oculta de los Puentes Cross-Chain