Hey there — want to state my explicit support. IMO this is a bigger deal than a typical ops improvement, and here’s why.
The current setup may quietly punish honesty. If an oracle member or DSM guardian suspects their hot key has leaked, the “right” move is an emergency governance cycle — roughly ten days of Aragon voting plus all the preps and comms — while the bad key keeps all its permissions. The best you can do in the meantime is ask for components to be paused around you. When disclosure costs that much, the rational move under uncertainty is to wait and hope. That’s how compromises get concealed — not malice, just process tax.
EDF fixes this at the root. The owner calls revokeDelegate() and the key stops working instantly; a replacement queues behind the cooldown and the seat stays online. Rotation becomes something you do on a schedule, not because something already went wrong — reactive flips to proactive. For example, the Chorus One incident last May needed an emergency vote and days of a known-bad key sitting on-chain. Under EDF: one transaction.
Timing matters too. Since late 2025 we have LLMs doing real security research — Anthropic’s red team saw agents go from exploiting ~2% of recent contract vulns to ~56% in a year. Automated probing of operator infra is trending toward free. “Key rotation requires a DAO vote” was already questionable; now it’s untenable.
One of my favorite parts of the proposal: terminate(). Even if the owner key gets compromised, the owner can still irreversibly shut down the whole delegation contract — execute() disabled, ERC-1271 failing closed. A role holder can finally cut itself off voluntarily, instead of the protocol pausing components around a party it can’t trust anymore. All in all, that’s a genuine attack-surface reduction.
One wish while we’re here: the same logic should eventually reach Beacon Chain validator keys. Today a leaked signing key can’t be rotated in place — the only path is exit and re-enter, queues and downtime included. That’s exactly the “wait and hope” incentive EDF removes, just one layer down. Out of Lido contributors’ hands, obviously, but worth mentioning.
Before the vote I’d want the per-seat cooldowns expectations and owner rotation recommendations nailed down. Neither changes my support — happy to see this go to implementation and audit.