Liquid's signers released 3,996 bitcoin because their own software said it was backed

The 11-of-15 multisig governs who may authorize bitcoin leaving the federation wallet. Whether a given request deserves authorizing is a separate matter, and Sunday's peg-out was authorized by exactly the parties meant to authorize it, with keys nobody stole ^{[4] [6]}. Liquid's own account is that the signers did what they were supposed to do, because the software told them the peg-out was legitimate ^[7]. Custody assurances written as thresholds speak to key compromise; the validity question sits outside that frame entirely.

Here is what a risk memo hears in "more than 80 federation members, 15 functionaries, 11-of-15 signing": fifteen independent parties who would all have to be wrong at once ^[5]. On the day, that bought fifteen parties reading the same claim from the same codebase, with the only dissenting nodes being the ones that had not taken last week's update and therefore stopped cold at block height 4,050,335 ^[11].

The 2019 provenance of the underlying flaw is the less useful half of Mononaut's account ^[9]. The half that should worry operators is the window: the exploited bug was days old and reached production through an ordinary upgrade, so the machines running current software were the exposed ones. On consensus code, a fleet-wide patch does not spread safety evenly; it puts every up-to-date node on the same answer at the same moment, right or wrong.

The arithmetic is worth doing by hand. Roughly 4,000 bitcoin valued at about $320m implies about $80,000 a coin ^[13]. The federation released roughly 3,996 bitcoin to an address the attacker controlled ^[17], of which 598.5 stayed with the attacker, a self-set fee of about 15% ^[12]. The public figures also do not close: 3,400 returned plus 598.5 retained is 3,998.5, two and a half coins more than the 3,996 reported released ^[18].

The control that could plausibly have caught this was a policy control, not a cryptographic one. Peg-outs on Liquid are meant to be restricted to authorized federation members and whitelisted Bitcoin destinations ^[14], and the freshly minted L-BTC nonetheless went out through SideSwap's peg-out service ^[17]. Wiz, a Bitcoin security researcher and Liquid member, argues that if SideSwap broke the policy limiting peg-outs to offline cold storage, it is partly responsible for the theft ^[15]. A destination list is a dumber mechanism than a cryptographic proof, and it does not care whether the proof verified.

Diligence on any federated bridge or n-of-m custody arrangement should ask two questions per signer instead of one. How many keys must an attacker hold to move the money, and what does each key holder consult before deciding to sign. When the second answer is one build of one codebase, the threshold in the first is counting machines rather than judgments, and the quorum collapses to 1-of-1 for any bug on the validity path. Time delays are the crude remedy other designs reach for, including the Drivechains and Rootstock proposals ^[16]. A delay would not have prevented the bad mint, but it would have given the nodes that rejected the transaction time to be heard before the coins left.