The loan and the lock applied to different assets
A token lock can keep a deposit in place without keeping the money used to buy it committed. X100 exposed that distinction on October 24, 2026: the first operation repaid its flash loan from value extracted through the treasury, while its voting deposits stayed frozen.
How two accounts became a majority
X100 was a token project on Ethereum, with a token called X. Its “Senate” was a project-specific governing body, not part of Ethereum’s governance. Registered accounts ranked by their current frozen X balances; the top three held seats, and two could approve treasury actions.
Bitquery’s on-chain investigation describes the first operation buying X with borrowed USDT, freezing it through two helper contracts, and using those seats to authorize a trade. A swap proposal could specify its own minimum return, which was set to zero. Execution did not require waiting after approval.
This made the distinction between accounts and people consequential. Two eligible accounts supplied two seats; the rules did not establish that they represented independent decision-makers.
What had to settle before the transaction ended
A flash loan provides temporary purchasing power. In the repay-in-full mode relevant here, principal and the required fee must be available before the transaction completes. If the conditions fail, the transaction reverts. Aave’s documentation also describes modes that can open a collateralized or delegated debt position, so immediate repayment is not a description of every Aave borrowing mode.
One transaction can call several contracts. For a successful repayment loop, an acquisition, a treasury operation and a sale can therefore happen before the lender collects what it is owed. The loan need not be repaid from the asset originally purchased.
Flash loans also support ordinary arbitrage and liquidations. The lender’s settlement rule tests whether it gets paid; that rule does not establish that another protocol has granted authority safely.
Follow the assets through the first operation
The figures below come from CertiK’s transaction analysis:
- Borrow: approximately 1.94 million USDT financed the operation.
- Acquire authority: two helper accounts each froze 19,000.000001 X, exceeding the previous second-ranked stake of 19,000 X by one millionth of a token. Each also paid a separate 100 X candidacy fee.
- Use authority: the majority authorized the treasury to exchange 400,000 X for about 35.68 USDT through liquidity controlled by the operation.
- Realize value: acquired X was sold into the legitimate market. The proceeds covered the lender’s repayment and a roughly 4.75 million USDT recipient payout.
The frozen deposits did not have to be recovered. For the financing to work, the accessible proceeds had to cover the obligations due at settlement, despite the stake remaining locked.
The thumbnail’s vault, chairs and padlock illustrate that relationship; they are not a transaction diagram or a photograph of the incident.
Why sacrificing the stake could still work
Consider a deliberately simplified example, not X100’s actual accounting. Someone borrows 100 units, spends them on a nonrefundable membership, and gains immediate authority over a club’s equipment. They use that authority to obtain equipment, sell it for 160 units, and owe 101 at settlement. Repayment leaves 59 units even if the membership can never be sold.
The arithmetic is 160 − 101 = 59. There is no second subtraction of the 100-unit membership purchase: its financing is already included in the debt repaid. But any additional fees or outside funding would still need to enter a complete profit calculation.
This explains why “the stake cannot leave” and “the acquisition cannot be financed briefly” are different propositions. A security review needs to compare the cost of authority with the value that authority can release, including a case where the deposit has no recoverable value.
A permitted swap can still give away value
SigIntZero’s analysis identifies another timing trap: the three-day proposal period set an expiry, rather than a minimum waiting time. A long-looking deadline did not prevent immediate execution once enough seats approved.
The economic check was similarly narrower than it appeared. Keeping enough X to cover recorded frozen balances protected that token liability. It did not independently require a reasonable return on a treasury exchange.
A pool follows its own trading rules. Uniswap’s pricing documentation explains why reserve-based execution needs an external reference when a contract must judge price fairness. Satisfying a pool’s invariant does not establish that the treasury received a sensible deal. Our liquidity-pool guide develops the reserve mechanics separately.
Keep the loss measures separate
Borrowed principal is temporary financing. A transfer of X is a token quantity. A recipient’s USDT payout is a realized payment. Adding all three would count overlapping stages as separate losses.
CertiK reports approximately 4.849 million USDT of net outflow from the legitimate pool, including two follow-on transactions. That is a broader measurement than the first recipient’s roughly 4.75 million USDT payout. Unsold X is excluded from that dollar valuation. The reviewed evidence does not establish a current recovery outcome.
Change the timing, the authority, or the permitted loss
OpenZeppelin’s governance guide describes historical voting snapshots and queued execution through a timelock. These address separate boundaries. A snapshot taken before an acquisition prevents the new balance from rewriting that checkpoint. An enforced delay puts execution outside the initial transaction. Neither by itself eliminates a takeover financed over a longer period.
These are design implications, not claims that X100 implemented a repair. A delay also needs monitoring and a usable response path. An independently enforced minimum return or loss limit addresses what a majority can spend; a proposal choosing its own meaningless minimum does not supply that protection. Price references need their own scrutiny, especially in thin markets.
The BonkDAO case shows another route from purchasable votes to treasury control. X100 adds a financing question to that review: can the assets reachable immediately after approval pay for acquiring the votes themselves?
