Compound Proposal 62 triggered erroneous COMP reward claims

Compound’s Proposal 62 took effect on Ethereum on September 29, 2021, changing how the lending protocol distributed its COMP governance token between suppliers and borrowers. That evening, the upgraded Comptroller began calculating outsized rewards for some accounts, creating claims far beyond the protocol’s intended incentive payments.

The clearest event-day evidence is on-chain. At 22:20:42 UTC on September 29, an account called Compound’s `claimComp` function and received exactly 29,665.158752987734887168 COMP from the Comptroller. The Ethereum transaction succeeded in block 13,323,324. Its logs show reward calculations across several Compound markets before the token transfer.

That transaction establishes what moved, when it moved and which contracts participated. It does not, by itself, establish the account holder’s identity, intent or legitimate reward entitlement.

What Compound acknowledged

Compound Labs subsequently acknowledged unusual COMP distribution following the execution of Proposal 62. Its event-day statement said the organization and community members were investigating discrepancies and that assets supplied to or borrowed from Compound were not at risk.

Robert Leshner, Compound Labs’ founder, attributed the problem to a bug in the new Comptroller implementation. His contemporaneous assessment placed an upper bound of 280,000 COMP on the amount that could be claimed improperly from the Comptroller. That figure was an incident-response estimate, not a verified final loss calculation.

Contemporaneous reporting also said the Comptroller held a limited COMP balance, while most tokens allocated for future rewards were held separately in a Reservoir contract. This distinction mattered: the bug did not create an unlimited supply of COMP, and the immediately exposed amount depended on the tokens available to the faulty distribution machinery.

Why the governance design mattered

Proposal 62 was intended to replace the existing arrangement in which a market’s COMP rate applied equally to suppliers and borrowers. Separate `compSupplySpeeds` and `compBorrowSpeeds` would let governance direct incentives toward one side of a market. The proposal also incorporated fixes for earlier reward-accounting problems.

The incident turned that routine-looking incentive upgrade into a governance stress test. Compound’s contracts were designed so that Compound Labs could not simply reverse the upgrade or switch off COMP distribution through an administrator-controlled emergency transaction. Leshner said any production change had to proceed through the protocol’s governance process, which required approximately seven days.

That constraint was simultaneously a decentralization feature and an operational limitation. It reduced the ability of one company to rewrite protocol state, but it also prevented an immediate repair while erroneous claims remained possible. On September 29, the central uncertainty was therefore not whether an anomalous transfer had occurred—the chain established that—but how much COMP could still be claimed before governance delivered a patch.

Separating the record from the estimates

No market-price calculation is used in this reconstruction. Contemporaneous publications attached dollar values to observed and potentially exposed COMP, but those estimates depended on a continuously traded token price and on assumptions about how much COMP would ultimately leave the Comptroller. The more durable measurements are the exact transaction amount and Leshner’s explicitly bounded 280,000-COMP scenario.

The available evidence also supported Compound Labs’ narrower statement that supplied and borrowed assets were not directly endangered. That did not make the incident immaterial. Erroneous distribution affected the COMP incentive pool, raised questions about upgrade review and exposed how slowly immutable governance systems could respond to faulty code.

Later context

Subsequent Compound governance proposals addressed the accrual error and its consequences. Those later remediation steps were not known outcomes on September 29, 2021, and are not used here to recast the event-day uncertainty as a resolved incident.