Balancer Governance Sets Formal Timeline for Protocol Wind-Down Following Major Losses
Balancer’s final act is to manage its own exit.
On September 29, a governance vote representing 17.2 million BAL—about 42 votes in total—set the process in motion to wind down one of DeFi’s earliest automated market makers.

The decision to sunset the protocol received near-unanimous support, with 99.2% in favor. A competing proposal to reboot the protocol under a different brand did not pass the same day.
Roughly eleven months earlier, a $128 million exploit stemming from a rounding error in a scaling function devastated Balancer’s v2 pools, and efforts to recover never fully overcame its impact.
In March, Balancer Labs formally dissolved its corporate entity, leaving a slimmed-down operation that incurred around $150,000 in monthly expenses, while August protocol income was under $30,000.
Starting October 30, most pools shift to a withdrawals-only state, though certain v3 pools may remain active until November 30 if requested by partners. Bug bounty protection ends October 30 for all pools.
BAL holders will need to wait until May 2027 before they can redeem their tokens for a proportional share of the remaining assets.
Behind this organized timeline, unresolved details remain: A post-mortem that leaves losses and recoveries not fully reconciled, a compensation structure requiring claimants to release Balancer from any further liability, and a voting mechanism error that initially blocked delegating token holders from overruling their delegates, which was later corrected and led to a re-vote.
The governance process determined how the protocol would close—but not a full accounting of the events that led there.
Who will ultimately verify the details of Balancer’s closure if the protocol writes its own history?
Credit: Unchained, Balancer, decrypt, The Defiant, Marcus Hardt, CoinTelegraph
For Balancer, the repercussions of November 3, 2025, never truly subsided.
The situation shifted from emergency response to legal management.
A fortnight after the exploit, Balancer published its official post-mortem, which did not fully match loss and recovery numbers across sections.
The opening section estimated the theft at $94.8 million, but the Financial Impact section cited $121.1 million in damages without clarifying the discrepancy of about $26 million.
The report listed approximately $45.7 million as either protected or recovered, although $19.3 million was liquidity untouched in paused pools.
The compensation plan covered about $8 million: This included $3.86 million recovered by external whitehats and $4.11 million internally secured from pools not yet compromised by attackers.
Recoveries managed by StakeWise for osETH and osGNO were valued at around $19.7 million in the reimbursement plan.
Claimants needed to sign off digitally.
Those seeking reimbursement were required to digitally consent to terms that released Balancer Labs, the DAO, the Foundation, and related parties from further liability related to the exploit. Whitehat bounty recipients, on the other hand, needed to complete KYC, ID checks, and sanctions screening.
To receive funds, claimants had to accept these terms.
Four months later, on March 23, co-founder Fernando Martinelli announced the dissolution of Balancer Labs, citing “real and ongoing legal exposure” from the security incident.
The requirement to release liability was part of the reimbursement process, and legal risk remained a driving factor for winding down.
Despite the company’s closure, the protocol remained online, and in April, token holders approved a streamlined operating plan, including a new operational company and a focus on self-sustaining finances.
Reducing expenses was one challenge; restoring user confidence was another.
How long can a “fresh start” last if it relies on dwindling resources?
A Slimmed-Down Operation
The April restructuring plan was comprehensive: it called for cutting costs, ending incentives, simplifying the token model, channeling all revenue to the treasury, and allowing v3 to drive potential growth.
Community members supported these measures. BIP-918 consolidated management under Balancer OpCo and reduced staff to 12.5 full-time equivalents from about 25.
BIP-919 ended BAL token emissions, phased out veBAL, redirected all protocol fees to the DAO (up from 17.5%), and introduced a BAL buyback at net asset value (NAV), capped at 35% of treasury assets, exercisable a year after the vote.
The estimated NAV per BAL was $0.16. Marcus Hardt, the CEO, later specified that the April buyback priced BAL at $0.1487.
As part of the overhaul, the v3 protocol share of swap fees dropped from 50% to 25%, with a clear warning that liquidity might leave if incentives were withdrawn.
By September’s end, DefiLlama data showed TVL at $58.5 million, down from $157 million in March.
The strategic overhaul did not reverse user outflows.
Several new products launched, such as AutoRange Pools, and integrations were explored. Hardt stated that v3 performed as intended.
“The product worked. It did not sell enough,” Marcus Hardt commented on Twitter.
Revenue statistics highlighted the problem. In October 2025—the last full month before the attack—protocol revenue reached $1.13 million.
By June 2026, monthly protocol revenue dropped to about $97,000.
For August, Hardt reported protocol revenue at $30,000, whereas DefiLlama listed $56,781 for the same month.
Most revenue continued to originate from legacy v2 pools.
Operating expenses still hovered around $150,000 per month, with treasury yield management by kpk returning about $25,000 monthly](https://forum.balancer.fi/t/bip-928-orderly-winddown-of-balancer-and-distribution-of-the-treasury/7107).
Hardt acknowledged a period of key personnel departures and reduced involvement, leaving the project under-resourced.
MAXYZ provided a more critical take. In a September 15 forum comment, the group claimed the development team quickly disbanded and lacked a concrete roadmap.
Hardt indicated that contributors received two months’ notice starting August 27, while the wind-down proposal became public 18 days later.
Holders would later vote on the protocol’s fate after contributors had already been notified of the restructuring. MAXYZ suggested an alternative route.
With the core team on notice, who would risk supporting ongoing operations?
Fork Proposal Fails
MAXYZ wasted no time.
Just a day after Hardt posted the wind-down plan, longtime participants Gosuto and Zekraken proposed a fork in the community forum.
On September 20, BIP-929 was submitted under the clear title “Fork and Reincarnate”.
Based on these assumptions, MAXYZ estimated annual revenue at $27.4 million—about 75 times Balancer’s current income.
The request was explicit: MAXYZ sought about 6 million BAL (3.5 million from the treasury, 1.6 million from the Labs fundraising safe, and 928k from the Labs team safe), valued at $690,000, as a pre-seed grant.
In exchange, the Balancer treasury would get 10% of the fork’s token supply, if such a token launched.
The proposal contained its own conflict: MAXYZ held two of the seven Treasury Council seats, which would execute the grant. MAXYZ offered to vacate both seats before any BAL was transferred, lowering the signing threshold from 5-of-7 to 4-of-5.
A forum participant estimated this would dilute other token holders by approximately 8.7%. MAXYZ agreed with this, quantifying the cost at a little over one cent per NAV per BAL](https://forum.balancer.fi/t/bip-929-fork-and-reincarnate/7112/5).
The proposal garnered support from both partners and internal contributors. Val of Rocket Pool advocated for postponing the pause and suggested Rocket Pool might move liquidity to a fork.
Both Royco and Tetu backed the fork idea, as did Danko from Balancer OpCo.
Xeonus, a former Foundation board member, declared his support and disclosed his role in founding MAXYZ and signing for its operations since July.
**The revised closure plan included a compromise: certain v3 pools could stay open until November 30 if requested by partners.
In the midst of voting, a technical issue surfaced.
On September 26, Fernando Martinelli shared signed messages from two holders with over 400,000 and 480,000 BAL, respectively, who reported zero voting power.
Hardt traced the problem to new voting strategies implemented July 7, which, unlike previous mechanisms, did not allow delegators to override their delegates. This limitation was not communicated to delegating holders.
MAXYZ subsequently corrected the strategy and restarted the fork vote.
Ultimately, the fork proposal failed, with 69.9% voting against and 17.33 million BAL participating.
With the alternative path rejected and the wind-down schedule in place, the question became: what would BAL tokens be worth at redemption?
Estimating Token Value
There were three values in play: market price, estimated treasury per token, and a final redemption value yet to be determined.
First, the market: BAL traded near $0.13 on October 1, shortly after the vote.
Second, Hardt’s estimate: A September 20 update put non-BAL assets at $9,959,416 against 63,068,821 redeemable BAL, using September 18 balances and DefiLlama prices, for an estimated $0.1579 per token. Hardt noted this was his own calculation, not an audited one.
He also updated the denominator, excluding all BAL held by the DAO and its entities as well as permanently locked BAL behind tetuBAL (about 9.3 million tokens).
The count of redeemable tokens included treasury BAL owed to tetuBAL holders, whose tokens are locked and cannot otherwise be redeemed. Voters later opted for the 50% distribution.
The actual redemption value will only be finalized and audited at the snapshot in late May 2027, with assets paid out in kind from the treasury’s holdings at that time.
That detail is crucial. Head of operations Danko advocated for converting to stablecoins so holders could know their exit value in advance.
Hardt, however, kept the plan for in-kind distribution. Fernando Martinelli supported this, citing the risk of selling partner tokens and the potential upside in a bull market.
The treasury is heavily weighted: Sentralis, a crypto risk firm, found staked AAVE made up 27% of assets as of September 21, noting the analysis was unsolicited.
Modeling possible outcomes for per-BAL value by May 2027, one model based on prior volatility placed 90% of results between $0.115 and $0.285](https://forum.balancer.fi/t/bip-928-orderly-winddown-of-balancer-and-distribution-of-the-treasury/7107/14).
A second model, replicating the year’s asset price history, projected a range of $0.103 to $0.206.
Additionally, holders will have six months (from late May to late November 2027) to burn their BAL for redemption; unclaimed shares will be automatically distributed among those who redeem.

tetuBAL holders, whose tokens cannot be unlocked, will receive treasury BAL equal to half the amount backing their tokens.
Voters favored the 50% allocation over 100%, by 12.2 million to 4.91 million BAL.
Regardless of the gap between BAL’s market price (around $0.13) and Hardt’s estimate ($0.1579), the redemption value will not be set until the May 2027 snapshot.
During the same week as debates over converting the treasury to stablecoins, the Treasury Council had placed a sell order for COW, a partner token.
What was the reason for this sale?
Unwinding Partnerships
The explanation emerged in another discussion: BIP-931, CoW DAO’s September 21 proposal to unwind its January 2025 token swap with Balancer.
On September 22, Danko revealed that the Treasury Council, in agreement with treasury manager kpk, had set up a TWAP order the prior week to sell COW tokens outside the BIP-770 swap agreement.
The Council canceled the order the same day Danko posted.
The cancellation followed CoW DAO’s proposal to unwind the January swap: Balancer would reclaim 200,000 BAL and about 29.95 ETH in exchange for its remaining COW, at a price set by a 15-day TWAP. Fernando Martinelli objected to a 30-day TWAP because it captured COW’s price surge. BIP-931 proceeded to Snapshot on October 2.
BIP-928 stipulates that wind-down-related asset movements wait for a community vote, while kpk recommends consolidating assets under its existing authority.
At publication, the forum did not clarify whether the COW sale was permitted under BIP-928’s voting restriction or under kpk’s mandate.
Unresolved issues remain.
Bug bounty coverage ends for all pools on October 30, even for v3 pools extended to November 30 by partner request. Hardt’s guidance to LPs in those pools was to withdraw by October 30 if they wanted to remain under coverage.
As of October 14, promised lists of extended pools, details on individual pool treatment, and withdrawal guides had not yet been published.
Redemptions are still over seven months away. Balancer has warned that any site or message offering early BAL redemption is fraudulent.
When the Treasury Council completes its final transfer in 2028, who will ensure these loose ends are addressed?
Balancer is executing the visible steps of a structured shutdown.
It provided notice, published a clear timeline, scheduled the treasury, kept withdrawals open, and explicitly warned holders not to trust unofficial redemption offers.
However, an organized closure only addresses process. The underlying post-mortem still contains unresolved numbers and a placeholder for final accounting.
The reimbursement structure required waiving liability.
A voting mechanism, active from July 7 to its September 26 reset, lacked the delegate override previously in place under BIP-921.
At publication, it was still unclear whose authority permitted the pre-vote COW token sale.
While the community vote determines the closure, these procedural details become part of Balancer’s historical record.
The wind-down process is scheduled to conclude with a final sweep in July 2028. The record will persist beyond the tenure of those responsible for addressing its unanswered questions.
When the final multisig is signed and the last operator steps away, who will be accountable for what never balanced?
Get new scam files the moment we publish them — usually 2–3 emails a week.