Balancer wind-down: BAL holders get paid in May 2027

BAL holders approved on September 29, 2026, with 99.19% of the 17,226,164 BAL voted, to shut down Balancer and distribute the treasury — and the first dollar will not be released until late May 2027. Balancer, the decentralized exchange protocol that introduced liquidity pools with configurable weights in 2020, reached a peak of 3,314 million dollars in custody, and by October 3, 2026, it had 58.4 remaining. The author of the proposal, signing as marcusblabs on the governance forum, measured the treasury on September 18 and published a per-token figure: 0.1579 dollars in non-BAL assets for each redeemable BAL, a self-measured and unaudited figure. On October 3, 2026, the market was paying 0.1321 dollars per BAL, 83.7% of that figure. We have re-read on-chain the Safe that concentrates 91.7% of that base, and the two large positions match the proposer's to the decimal; what has changed is the price, because 34.8% of that Safe is in a single token (AAVE) which rose 28.6% between September 18 and October 3, 2026. This article breaks down exactly what the BAL holder receives, in what form (in kind: the existing tokens, not dollars) and against what schedule, and places it alongside the other three protocols that announced their closure between September 4 and October 2, 2026: Router Protocol, Blast, and Harmony. Of the four, Balancer is the only one returning treasury to the token holder.

What did Balancer approve on September 29, 2026 and what does the BAL holder receive?

Proposal BIP-928, "orderly closure of Balancer and distribution of the treasury," was voted on Snapshot from September 25 to 29, 2026. The ballot featured four options instead of three, as the vote itself decided a specific parameter: what fraction of the BAL permanently locked in Tetu —a yield aggregator that locked its veBAL with no possible exit— is compensated from the treasury. Combining the two affirmative options, 17,086,591 BAL voted in favor versus 139,573 against, with 42 voters. The variant setting that fraction at 50% received 12,179,932 BAL, while the 100% option reached 4,906,659, thus setting the parameter at 50%.

During the same period, the alternative BIP-929, "Fork and Reincarnate," which proposed keeping the infrastructure alive under a different name, was voted on. It was rejected: 12,113,520 BAL against versus 5,213,625 in favor, with 51 voters, the highest number of participants among the last twelve proposals in the Balancer governance space.

What the BAL holder approved is a three-part package that should be separated, as each has its own timeline. The first is the product shutdown: on October 30, 2026, the pools (the common token funds against which trades are executed) that can be paused will switch to withdrawal-only, and in those that cannot be paused, the protocol fee will drop to zero where contracts allow. Partners needing more time for a v3 pool had to request it before October 16; those pools remain active until November 30, when the v3 Vault is paused. The second part is the organization's closure: the notice period for contributors ends on October 31, and starting November 1, a total wind-down budget of 400,000 dollars will be in effect. The third is the distribution.

The distribution works as follows. Starting in late May 2027, each holder burns their BAL against a claim contract and receives their proportional share of the treasury in the tokens the treasury holds at that time, not in dollars. The window lasts six months, until late November 2027. The proposal then sets a round two within two months after that window closes —a maximum timeframe— which no longer requires claiming anything: it is an automatic airdrop to the addresses that redeemed in round one, in proportion to what each one redeemed, and distributes three things: the unspent closing budget, whatever has reached the treasury since then, and the portion that no one claimed in round one. Six months later, there is a final sweep of whatever has appeared since, to the same addresses and in the same proportion. If both deadlines are exhausted, that sweep falls in mid-2028; the proposal does not publish the date, so that is our own estimate based on their schedule. Anyone who does not redeem in round one is excluded from the two following rounds.

There are two more rules that change the calculations for anyone looking at the price chart and doing the math. BAL is not a distributable asset: it is burned, and what you receive remains outside of BAL and does not convert back into BAL. Furthermore, the BAL held by the organization itself and its entities is not redeemable, nor is the BAL behind Tetu's permanent lock, which can never become BAL again.

What does the token holder receive in Balancer, Router, Blast, and Harmony, and when?

Between September 4 and October 2, 2026, a four-week period, four protocols announced they were shutting down. All four are classified as "closures," yet the token behavior differs in each case. Here is what the holder of each of the four tokens receives, when, and in what form:

The rows for Router and Harmony share a limitation that should be disclosed. In both cases, the announcement is an X article, a format whose body is not returned by the X API: for Router, we know it was published on September 4, 2026, at 23:15:40 UTC, and for Harmony on September 6 at 14:21:24 UTC, but the cells for these two rows are built on what The Block and The Defiant wrote, not on the original text. Only Balancer and Blast are backed by a readable document from the protocol itself. What is verified through other means is the magnitude of the Router burn: 303,333,198 ROUTE is exactly the difference between the total supply published by CoinGecko, 982,072,351.30 ROUTE, and the circulating supply, 678,739,153.30. This token-for-token match supports the announced figure, and it is worth highlighting what it measures: on October 4, 2026, at 14:31 UTC and again on October 5 at 20:49:40 UTC, the total supply of ROUTE remained 982,072,351.30 — unchanged in both readings and with the difference against the circulating supply fixed at 303,333,198. The Defiant describes these tokens as pending in the treasury and states that Router did not identify addresses, a mechanism, or an execution date. The burn has been announced; it has not been executed in the public data.

In all four cases, the token price moved on the day of the announcement, and here the timing matters because the movements last for hours, not days. ROUTE was trading at 0,0001117 dollars at the 23:00 UTC mark on September 4, immediately preceding the announcement: by 09:00 on the 5th, it was at 0,0000504 —a 54.9 % decrease— and it closed September 5 at 0,0000682 —39.0 % below the pre-announcement point. BLAST was worth 0,0004103 dollars at 14:00 UTC on October 2, forty-five minutes before the publication; in our first reading on October 3, it was 0,0002203 (a 46.3 % drop) and in the second, at 19:57 UTC, 0,0001696 — a cumulative 58.7 % decline in the twenty-nine hours following the announcement. ONE did the opposite: from 0,0007450 dollars at 14:00 UTC on September 6 to 0,002512 at the 13:00 UTC mark on October 3, a 237.1 % increase, in the only one of the four cases where the token survives one-to-one; our second reading that day, at 19:57 UTC, placed it at 0,002638. And BAL jumped 32.8 % in two hours on the night of September 26, from 0,1213 to 0,1611 dollars, once the direction of the vote became clear.

If you are interested in the other type of closure —exchanges that shut down their order books and where their volume ends up— that is measured in the six exchanges that closed between June and July 2026 and in the BitMEX shutdown.

How much is the Balancer treasury worth read on-chain on October 3, 2026?

It is worth knowing the scale of this protocol's fall before accounting for what remains. The DefiLlama daily series, as of October 4, 2026, provides the full trajectory:

On October 3, 2026, the protocol retained 1.8% of its peak. The August 31, 2026 exploit, which funds proposal BIP-930, hit a protocol that had already lost 98% of its liquidity; the November 2025 event is detailed in the 2025 crypto security report and in the review of bridge architectures after the attacks.

The proponent of BIP-928 measured the treasury on September 18, 2026 —Ethereum block 26,005,414, using DefiLlama prices from 14:00 UTC that day— and created the document with the methodology and all addresses at 19:39 UTC. The figure was not published on the governance forum until September 20 at 21:30 UTC, in a message where it is labeled as a "self-measured, unaudited" figure. It provides a distributable base of 9,959,416 dollars in non-BAL assets and 63,068,820.84 redeemable BAL, which amounts to 0.1579 dollars per BAL at that day's prices.

We re-examined on-chain, on October 3, 2026, the Safe that concentrates 91.7% of that base: the Treasury Safe managed by kpk —formerly karpatkey, the DAO treasury manager— as the treasury administrator for Balancer. The two large positions match those of the proposer to the decimal: 19,549.3993 stkAAVE and 350.8384 weETH are identical, and the USDC vaults and the USDC lent on Compound align with the expected accrual over those two weeks. What changes is the price. This is the breakdown:

The main Safe is worth 11.4% more than on September 18, 2026, an increase of 1,040,029 dollars. The cause is almost entirely in one place: AAVE went from 140.66 to 180.93 dollars, a 28.6% rise, and the treasury holds 19,549.3993 stkAAVE (AAVE deposited in the Aave protocol safety module). This position alone explains about 787,000 dollars of the increase and accounts for 34.8% of the Safe. The rest consists of four stable positions —USDC in two managed vaults, USDC lent on Compound, and sDAI on Gnosis—, a bundle of weETH, and small tails of COW, the SAFE token, HOPR, and POL.

The small tails are precisely where the amounts do not add up, and the reason is that the treasury manager is already selling them. Between September 18 and October 3, 2026, the Safe ran out of rETH (the 62.6945 it held, about 196,700 dollars at October 3 prices) and sGHO (150,704 GHO, about 150,400 dollars); the SAFE token (the governance token for Safe, the multi-signature wallet contract used by the treasury itself) dropped from 845,094 to 453,436.19 units, a 46.3% decrease, equivalent to about 43,000 dollars at the October 3 price; HOPR dropped from 617,205 to 264,516.49 — a 57.1% decrease, about 4,700 dollars; and COW was cut from 1,230,445 to 1,179,455.73 — a 4.1% decrease, about 8,200 dollars. Conversely, the main USDC vault gained 317,735 USDC, and 81.1355 native ETH appeared where there was previously no ETH line item. The Safe total remains higher than that of September.

The re-reading on October 4, 2026, at 14:26 UTC, twenty-five hours after the first one, reveals the operation: the SAFE token drops by another 27,975.53 units and the loose USDC rises by 3,131.80 — at 0.112164 dollars per SAFE, both figures represent the same sale. Everything else in that Safe remains identical, including the 19,549.3993 stkAAVE, the 350.8384 weETH, the 1,179,455.73 COW, the 264,516.49 HOPR, and the 81.1355 native ETH, with the only variations being the interest accrual from the Compound and Aave positions. This is what point 7 of BIP-928 mandates before round one: consolidating small and illiquid positions so that the claim fits into about six tokens. It is already happening, token by token, before any of the October deadlines expire.

The two positions upon which the calculation hinges also have a corroboration that does not depend on us or the date we read them: the 19,549.3993 stkAAVE and the 350.8384 weETH are the same as those measured in the proposer's document on September 18, fifteen days prior, and the ETH balance of the DAO's multisig is the same as that published in the BIP-930 table from September 16, seventeen days prior. These are three measurements taken by two different parties on three different dates with the same result.

What does each figure in Balancer's closing measure: distributable base, valued treasury, and budget?

Six fields of the Balancer closure resemble each other and measure different things. Each with its literal label:

• "Distributable base, non-BAL assets": what goes into the distribution. It amounts to 9,959,416 dollars according to the proposer at prices from September 18, 2026. It excludes BAL and everything settled in BAL, and is already net of the wind-down budget.
• "Valued Treasury": the total value of everything the organization holds at a given time. This is a higher figure and is not distributable as is, because it includes 3,435,806.35 BAL from the treasury itself which are excluded by rule, funds belonging to third parties, and the budget that the wind-down still needs to spend. Neither the proposal nor the proposer's document publishes a total for this field: what they publish is the distributable base and the items excluded from it, one by one. The DefiLlama treasury page does not serve as a substitute either, because its breakdown by token stopped on April 30, 2026.
• "Wind-down budget": 150,000 dollars from November 1, 2026, to May 2027, 30,000 from then until the final sweep, and a 220,000 reserve only to be touched if necessary. Total 400,000 dollars, and it is a cap, not a transfer: first, what is already in the operational account—which had 763,010 dollars on September 18—is spent, and whatever is not spent returns to the distribution in round two or the final sweep.
• "Redeemable BAL" (the denominator): 63,068,820.84 BAL. This is the total supply, 72,373,883.13 BAL, minus all BAL held by the organization and its entities, minus the 2,786,085.71 BAL behind Tetu's permanent lock, plus 1,395,593.07 BAL that the treasury pays to tetuBAL holders—the liquid receipt Tetu issues for its veBAL, the locked BAL that grants voting power. About 9.3 million BAL do not redeem, and that increases the value per token for those that do.
• What remains excluded as belonging to others: funds recovered from protocol attacks, which belong to the liquidity providers of the affected pools. The 296.401711 ETH recovered from the August 31, 2026 exploit is the easiest case to mistakenly include.
• A conditional line: the 600,000 BAL lent to Amber Group, the trading desk that took them in 2024 under BIP-664, are excluded as long as they are owed to the treasury. If they have not returned at the time of measurement, they count as circulating and are redeemable.

The proposal also establishes which measurement is valid, and it falls outside the previous ones: the base is measured and audited at the block where round one opens, announced at least two weeks in advance, at the end of May 2027. Everything published before that date sets the order of magnitude and the method; the figure that settles is the one from May 2027.

Why are the 296.401711 ETH recovered from the Balancer exploit not for the BAL holder?

On August 31, 2026, an attacker exploited a rounding flaw in the single-sided entry flow of Balancer V1 pools, immutable code deployed in 2020 with no pause mechanism. The technique was replicated across 120 pools, draining 1,393,694.67 dollars at the prices at the time of the attack. White hat actors front-ran some of the copies and, according to proposal BIP-930, the primary attacker partially returned funds.

The refunds arrived in five transactions to the organization's multisig on Ethereum —the multi-signature address controlled by the DAO— between September 8 and 16, 2026: 120 ETH; 52,061; 100,2191; 15,948; and 8,1736. Total, 296,401711 ETH. We read that multisig on-chain on October 3, 2026, and again on October 4, twenty-four hours later: both readings show the same native balance, matching the total from the BIP-930 table published on September 16 up to the sixth decimal place: 296,401711 ETH. Beyond that decimal, there are 9×10¹⁰ wei of execution dust remaining, so the match is to the sixth decimal and not to the wei.

That figure, at October 3, 2026 prices, represents 794,872 dollars. Compared to the 1,393,694.67 dollars drained, it yields 57.0%, and this percentage is our own calculation and has a caveat: the numerator is valued at the October ETH price and the denominator at the token prices on the day of the attack. Since they do not share a price basis, the 57.0% serves as an order of magnitude; an audited recovery rate would require valuing both legs on the same date. To the loss, one must add that approximately 169,000 dollars were consumed in payments to block builders during the race to intercept the funds.

And what matters for the distribution: that money remains outside the BAL holder's reach. Point 8 of BIP-928 states this, and the proposer's document does not include it in the base. It belongs to the liquidity providers of the affected pools, distributed among them in ETH, pro rata per pool according to how much was drained from each, and pro rata by holding within each pool, with the snapshot taken at the last block prior to the first transaction of the attack. The multisig also holds 177,5951 WETH, 80,2779 wstETH, and 78,0039 rETH, which at October 3, 2026 prices are 989,024 dollars, plus 12,518 USDe: around one million dollars that also does not enter the distribution to holders. Anyone looking at the balance of that multisig and adding it to the treasury inflates the distribution base by around 18%. The four positions remain identical in both readings, those of October 3 and 4.

Why is the market paying 83.7% of the BAL value calculated by the proposal?

BAL was trading at 0.1321 dollars on October 3, 2026. With 69.83 million BAL in circulation, that represents a market cap of 9.22 million dollars. There are three live value figures per token, plus one historical reference, each with its own basis:

The three distribution figures remain above the market price, and both of our figures are higher than the proposer's. We are not choosing just one: the $0.1612-$0.1750 range exists because we have not been able to rule out that part of the $397,709 from the v3 buffers has returned to the main Safe, in which case our repricing of the remainder would be double-counting up to $400,000. Therefore, $0.1612 is the safe lower bound and $0.1750 is the ceiling of the exercise.

The three percentages in this calculation measure three different things and should be given their own names, as they are easily confused. The 83.7% is a ratio: the market price of BAL divided by the proposer's value per token. Its complement, 16.3 percentage points, is what is known as the discount, and in dollars, it is a subtraction of 0.0258 dollars per token. And the 31.2% is a third thing, neither a price ratio nor a subtraction: it is an annualized rate, our own calculation, which equates 0.1321 dollars today with 0.1579 dollars in 240 days, which is the time remaining from October 3, 2026, to May 31, 2027. Based on the other two figures in the range, that same annualized rate would be 35.4% and 53.4%.

The discount has narrowed without closing, and the path was not a straight line. With BAL at the $0.1114 that the proposer's own document uses —DefiLlama price at 14:00 UTC on September 18, 2026— the ratio against their figure was 70.6%. In the following eight days, it rose to 76.8%, with BAL at $0.1213 at the 19:00 UTC mark on September 26. Between that point and 21:00 UTC on the same day, a two-hour window, BAL jumped 32.8% to $0.1611 and the ratio reached 102.1%: during that hour, the market paid above the per-token value calculated in the proposal. It later receded, and with BAL at $0.1321 on October 3, the ratio is back at 83.7%. The net result is 13.1 percentage points higher than on September 18, distributed across three segments of +6.2, +25.3, and −18.4 points. A nuance in the data: the starting point is a DefiLlama price and the other three are from CoinGecko, which for that same September 18 at 14:00 UTC gives $0.1119 and a ratio of 70.9%, three-tenths of a point higher. Two days after the close of that series, on October 5, 2026, at 20:49 UTC, BAL was at $0.1361: the ratio against the proposer's $0.1579 rises to 86.2% and the discount drops to 13.8 percentage points, $0.0218 per token. The complete ratio series stands at 70.6% → 76.8% → 102.1% → 83.7% → 86.2%. The valuation tables in this article remain anchored to October 3 on purpose: moving the date would require revaluing the entire Treasury Safe, position by position. At the October 3 market price, the 63,068,820.84 redeemable BAL are worth $8.33 million against a basis of between $9.96 and $11.04 million.

What remains in that gap are three specific things, and the inflated treasury is left off the list because the numbers add up. There are eight months of waiting, during which the treasury can gain —the proposal states that the treasury yields about 25,000 dollars per month the way kpk manages it— and can lose due to price, because 34.8% of the main Safe is in AAVE. There is the payment in kind: the claim consists of about six tokens and each recipient must sell them themselves, which the proposer himself acknowledges as expensive for the holder and for auditing. And there is the measurement risk: the applicable base is that of the opening block in May 2027, after costs, confirmed third-party funds, and any recoveries that come in. The figure of the discount on asset value also appears in bitcoin treasury companies trading below their net asset value; the operational difference in Balancer is that there is a published delivery date.

How much participation did the Balancer open votes have on October 5, 2026?

Two proposals remained open on October 5, 2026, and close on the 6th at 18:00:00 UTC. Both involve funds that affect the distribution.

BIP-930 approves the distribution of the 296.401711 ETH recovered from the exploit to the liquidity providers of the 120 affected pools, including the claim mechanism, the table per pool, and the address lists. BIP-931 unwinds the token swap that Balancer and CoW DAO executed in January 2025 —200,000 BAL for 631,850.824075 COW with a two-year lock-up—: Balancer returns its 1,179,455.7313 COW, recovers the 200,000 BAL, and receives approximately 30.22 ETH for the remaining 547,604.91 COW, based on a fifteen-day time-weighted average price. That average price implies COW at 0.1480 dollars, 8.5% below the 0.161744 at which it traded on October 3, 2026, and 6.6% below the 0.158406 recorded on October 5 at 20:49 UTC. The reason given by the proposal is operational: distributing an illiquid governance token among thousands of redeemers forces each one to sell, and holding it for eight months exposes the treasury to price risk without a strategic rationale. If BIP-931 passes, the proposer's value per BAL drops from 0.1579 to 0.1570 dollars, a 0.6% decrease at homogeneous September 18 prices, because more value leaves than enters, but 200,000 BAL are also removed from the denominator. The amount of COW that the proposal mandates to be returned, down to the last of its ten decimals, is exactly what the Safe held on October 3 and 4, 2026.

The status of those two votes on October 3, 2026, three days before closing, was as follows: both had one voter and 8.54 BAL each, compared to a quorum (minimum participation for a proposal to be valid) of 5,000,000 BAL, as established by BIP-924. This represents 0.00017% of the required minimum. Two readings taken six hours apart on the same day yielded the exact same figure down to the fifteenth decimal place, confirming that those 8.54 BAL were not a partial refresh of the voting service.

Two days later, on October 5, 2026 at 20:51 UTC, both proposals remained open and had recorded four voters and 4,643,380.955163 BAL, representing 92.8676% of the quorum, just 356,619.044837 BAL away from the five million mark, with all weight behind the favorable option: the breakdown by options published by Snapshot shows 4,643,380.955163 in favor and zero for the other two options in both cases. Neither of the two readings indicates who wins; together, they illustrate how the quorum is formed in this protocol.

Snapshot publishes every vote with its issuer, its weight, and the exact time it was entered, so the 4.64 million BAL can be broken down one by one. There are four votes:

The first two addresses total 4,643,343.321592 BAL, representing 99.99919% of the votes cast, and both votes were submitted less than twenty-seven hours apart. These are not equivalent positions, and the census of previous proposals reveals how they differ. The fourth address is the single voter who carried both proposals on October 3: their vote was cast on October 2 at 23:34:52 UTC, they retain the same 8.544912 BAL, and they now account for 0.00018% of the total. Their absolute weight remains the same as in the first reading; what changed was the denominator.

Whether a reading of 8.54 BAL three days before the close is alarming or routine depends on how things are approved in Balancer, and the series of the last twelve proposals in its governance space provides the benchmark:

Balancer's procedural proposals pass with between seven and eleven voters and, except for the one that revoked Aura's voting power, with between 6.45 and 6.56 million BAL, just above the five million quorum. Three of them passed with exactly 6,516,516 BAL and seven voters, and the two major ones, the closure and its alternative, moved 17.2 and 17.3 million BAL with 42 and 51 voters.

Who provides the quorum in Balancer and why does their weight change from one vote to another?

The voter census for these proposals is also published on Snapshot, and reading it turns an inference into a measurement. In BIP-925, one of the three that concluded with exactly 6,516,516 BAL, the two largest voters are 0x9f74662a… with 2,823,304.966896 BAL and 0xa9B3C2… with 2,127,125.831389: the same two addresses that on October 4 and 5 provided 99.99919% of the votes in BIP-930 and BIP-931. The small voter, 0x1b35fcb5…, was also in BIP-925, with 8.564540 BAL. In the two large ones, the census order is different: in BIP-928 and BIP-929, the largest voter is 0x0644141D…, with the same 2,784,158.350628 BAL in both, 0xa9B3C2… is second with 2,164,123.346969 — and 0x9f74662a… appears seventh in BIP-928 with 616,128.195822 and twenty-ninth in BIP-929 with 135,291.621131.

From this, two measured facts emerge, and neither reveals who is behind any specific address. The top voter by weight for procedural proposals and the top voter by weight for the closing are different addresses. Furthermore, the two addresses that held the quorum for BIP-930 and BIP-931 at 92.87% on October 5 are the same two that led the census for BIP-925.

In the census, something stands out that requires explanation: the voting weight of 0x9f74662a… shifts from 2,823,304.97 → 616,128.20 → 135,291.62 → 2,521,359.36 BAL from one proposal to the next. The explanation is precise, and the clearest examples are BIP-928 and BIP-929, the two closing proposals, because they were voted on the same Ethereum block, 26,056,112, and with the same set of eight counting strategies: the BAL balance across six chains, a call to the votingPower method of a voting power contract, and a strategy called delegation on that same contract, which adds the weight delegated by other addresses to the voter.

With the block and strategies being equal, an address should have the same weight in both, and that is exactly what happens: of the forty-one addresses that voted on both proposals, forty have identical weight digit for digit. 0x9f74662a… is the only exception, losing 480,836.574691 BAL when moving from one to the other. BIP-929 had ten voters that BIP-928 did not have —51 versus 42— and one of them, 0x91B9E59614995E13a32e36440Ac524825F7AE39E, voted with 480,836.574691 BAL. It is the exact drop:

Having tested all combinations of up to four of those ten new voters, that is the only one that fits, and it fits perfectly. This is the documented behavior of the delegation strategy: it counts the delegated weight within the delegate's vote and removes it as soon as the delegator votes on their own in that same proposal. Thus, the weight of 0x9f74662a… does not move because the address bought or sold BAL between one vote and another: it moves because, in one of the two, the person who had delegated to them voted directly.

Where the measurement ends, because the boundary matters. Read from Snapshot: the shared block, the strategy game of the two proposals, the two weights of 0x9f74662a…, the drop, the vote of 0x91B9E59614… and the coincidence between the drop and that vote —exact to the six decimal places published by the API, with a 6×10⁻¹¹ BAL difference below, which is noise from the floating-point format in which the weights are returned—. Not read: the on-chain delegation record. That 0x91B9E59614… delegates to 0x9f74662a… is not confirmed via that route; what is confirmed is that the figure matches to the decimal and that the delegation strategy produces precisely that effect.

For anyone reading the census, the consequence is that the two addresses that held a 92.87% quorum on October 5 behave differently. 0xa9B3C2… carries the exact same 2,164,123.346969 BAL in BIP-928 and BIP-929 down to the last decimal: its weight does not depend on more people voting. That of 0x9f74662a… does. Part of the concentration of Balancer governance, then, is not holding but aggregation: the same BAL can appear as one large vote or as several smaller ones depending on who bothers to vote, and a census of seven voters and another of fifty-one may be counting overlapping power. What each address voted, when, and with what weight is public; why it delegated or voted is not revealed by any of these figures, and the mechanics of a counting strategy do not constitute behavior.

That explanation does not extend to the other comparisons, and it is worth noting where it stops. BIP-930 and BIP-931 are voted on at block 26,106,075, which is 49,963 blocks after the two closing ones, using the same eight strategies. BIP-925 was voted on at block 25,802,574 —303,501 blocks before the two open ones— and with seven strategies instead of eight, the first of which was a different variant of the count with delegation. Between those blocks, there are weeks during which an address may have bought, sold, locked BAL, or received and lost delegations, so the weight differences compared to BIP-925 are not explained by the mechanism above and we do not force them.

So the quorum of this protocol is decided by a handful of addresses —and part of that handful is aggregated weight from third parties, not their own holdings—, and the jump from October 3 to 5 is that mechanism in plain sight: two votes cast less than three days before the closing cover 92.87 % of the minimum requirement. The verifiable consequence is that the governance tasked with executing this closure depends on those addresses continuing to vote until the end — and the proposer himself counts fourteen months from his message on September 20, 2026, until the closing of round one at the end of November 2027, with the final sweep even further away. If you want the general framework of how a DAO makes these decisions, it is in what is a DAO; what happens when contributors leave before the voters is measured in the Aave governance crisis. And the net value buyback that BIP-919 approved in April 2026 and BIP-928 canceled belongs to the family of mechanisms compared in the review of fee switches and buybacks in DeFi. Another on-chain voting model, with its own quorum mechanism, is described in how to vote on-chain in Solana.

What changes on October 30, 2026, in Balancer pools?

For the liquidity provider, and not for the BAL holder, the key date is October 30, 2026. On that day, the pools that support pausing will be paused and switched to withdrawal-only, with recovery mode activated where the contract requires it to ensure the exit remains open; those that do not support pausing will continue to function with the protocol fee set to zero where the contracts allow. The bug bounty program ends that same day. On November 30, the v3 Vault will be paused, and from then on, only a simplified withdrawal interface, necessary indexing coverage, and documentation will remain, maintained until the veBAL locks expire and throughout the two distribution rounds. The contracts are non-custodial: withdrawing does not depend on Balancer remaining operational, which is the part that distinguishes this closure from that of a centralized platform.

Liquidity is already moving, and measuring it on October 3, 2026, requires specifying which data point is being read. DefiLlama's daily series shows 65.32 million dollars on August 30, the eve of the V1 exploit; 58.90 million on September 15; 58.33 on September 29, the day the vote closed; and 58.40 million at the daily mark on October 3. The live aggregate for that same October 3, however, shows between 49.57 and 49.85 million, and the gap lies entirely within one leg: v2 on Ethereum dropped from 24.81 million at the daily mark to 16.00 million in the live aggregate, a decrease of 8.81 million, while v1, v3, and v2 on other chains remained flat. The consolidated daily mark for October 4 shows 49.41 million for the protocol and 15.64 million for v2 on Ethereum: the outflow stayed where it was and was not a short-term fluctuation. If you are in one of those pools, cómo funciona una pool de liquidez and cómo comprobar el estado de un protocolo cover the mechanics of the exit.

The closure does not stem from an exhausted treasury either. The proposal publishes its own accounts: Balancer's monthly spending was approximately 150,000 dollars, protocol revenue dropped from about 97,000 dollars in June to around 30,000 in August 2026, with the majority coming from v2, and the entire closure costs 400,000 dollars. The proposer's argument is one of distribution: continuing on the current path spends the treasury to reach the same destination later, and that treasury belongs to BAL holders.

What remains to be verified in the Balancer closure before May 2027?

Each date on the calendar approved in BIP-928 is a possible check:

Three dated checks summarize what remains open. The first is October 30, 2026: when the pools switch to withdrawal-only, nothing will be collectible yet, so if BAL is trading then at 95% or more of the value per token, the 16.3 discount points from October 3 —13.8 on the 5th— were not the price of waiting and another explanation had to be sought. The second is the denominator: the audited census of May 2027 is announced two weeks in advance and can be compared line by line with what is published today; if the amounts deviate by more than 10% at equal prices —due to costs exceeding 400,000 dollars, third-party claims, or leakages— the base was not what was measured. The third is the liquidation: if in the week prior to the opening of round one the ratio between the BAL price and the value per token does not exceed 95%, that is, if more than five discount points remain, what separated the market from the treasury was basket illiquidity, redeemers who do not appear, or tax friction, and not time.

For the BAL holder, what can be evaluated now that could not be before September 20, 2026, fits into four data points: the figure published by the proposer and their method, an on-chain re-reading that confirms it in the two large positions and increases the value of the main Safe by 11.4%, the exact date and method of collection, and the fact that of the four protocols that announced their closure between September 4 and October 2, 2026, only one is distributing its treasury. What none of those four things indicate is whether a gap of 0.0258 dollars per token on October 3, or 0.0218 two days later—eight months out and based on a basket to be re-measured in May—is a lot or a little. That calculation is for each individual to make.

Sources and links: BIP-928 on Snapshot (orderly winddown, result and text) · BIP-929, "Fork and Reincarnate" (rejected), with its snapshot block, its eight counting strategies and its census of 51 voters · BIP-930, distribution of recovered funds from the V1 exploit · BIP-931, Balancer and CoW DAO · BIP-925, one of three procedural proposals with 6,516,516 BAL and seven voters, with its voter census · governance forum thread for BIP-928, with the updated figure per BAL from September 20 · method and addresses for treasury measurement · BIP-930 on the forum, refund table · Balancer Treasury Safe on Blockscout · Balancer DAO multisig on Blockscout · DefiLlama, Balancer TVL · CoinGecko, BAL price and supply · official Blast post regarding its shutdown · Router Protocol post regarding its shutdown (X article from Sep-4-2026) · Harmony post regarding the shutdown proposal · The Block on the Router Protocol shutdown · The Defiant on the Router Protocol shutdown · The Block on the Harmony shutdown proposal