Ethereum's Glamsterdam upgrade — the network's most ambitious hard fork since the Merge — was pulled from its final testnet four days before activation after security researchers identified a class of gas-accounting vulnerabilities that could render failed operations permanent in at least 13 depl...
"Already found 13 real-world examples." — ChainSecurity, Ethereum security research firm, responsible disclosure statement, October 2026
Ethereum's Glamsterdam upgrade — the network's most ambitious hard fork since the Merge — was pulled from its final testnet four days before activation after security researchers identified a class of gas-accounting vulnerabilities that could render failed operations permanent in at least 13 deployed smart contracts. The Ethereum Foundation confirmed the postponement on October 22, 2026, following the ACDE #248 core developer call. No replacement date has been set.
Glamsterdam packages enshrined proposer-builder separation (EIP-7732), block-level access lists (EIP-7928), multidimensional gas metering, and a proposed 3.3x gas-limit increase from 60 million to 200 million. The upgrade had cleared its Sepolia testnet activation on October 6 without incident. But ChainSecurity's October 13 disclosure of insufficient gas griefing risks in EIP-8037 and EIP-8038 forced developers to abandon the Hoodi testnet launch scheduled for October 26 at epoch 132,352. The delay puts Glamsterdam's previously discussed early-December mainnet window under direct pressure. Developers had entered the ACDE #248 call expecting to choose between December 1, 8, or 9 for mainnet activation; instead, they voted to halt.
The postponement arrives at a fragile moment for Ethereum. ETH dropped 3.5% to $2,386 on October 22, the day of the announcement, amid concurrent spot ETF outflows of $110.6 million. The network's 863,000 active validators — down from 975,000 at year-start — and 43.8 million staked ETH ($105 billion at current prices) underscore the scale of assets dependent on safe protocol upgrades.
Glamsterdam merges the Amsterdam execution-layer fork with the Gloas consensus-layer fork. It carries at least eight EIPs targeting three structural problems: block-production centralization, gas-pricing inaccuracy, and throughput limits.
| EIP | Layer | Function | |---|---|---| | EIP-7732 | Consensus | Enshrined proposer-builder separation (ePBS) | | EIP-7928 | Execution | Block-level access lists for parallel transaction execution | | EIP-8037 | Execution | Separate gas reserve for state creation | | EIP-8038 | Execution | Repriced access and modification of existing state | | EIP-8011 | Execution | Multidimensional gas metering | | EIP-8032 | Execution | Size-based storage pricing | | EIP-2780 | Execution | Resource-based intrinsic transaction costs | | EIP-8261 | Consensus | Epoch-based gas-limit schedule |
The headline feature, EIP-7732, writes proposer-builder separation directly into Ethereum's consensus rules. Currently, block proposers outsource construction to specialized builders through off-protocol relays — principally MEV-Boost. That system works but inserts trusted intermediaries into block production. As of October 2026, three builder operations (Titan at 52.16%, BuilderNet at 24.63%, Quasar at 15.06%) produce approximately 92% of MEV-Boost blocks. ePBS would make relays optional rather than required, settling builder-proposer payments natively on the consensus layer.
EIP-7928 introduces block-level access lists, enabling transactions that do not touch the same state to process simultaneously rather than sequentially. Combined with the gas-limit increase to 200 million (from the current ~60 million), these changes represent a 3.3x theoretical capacity expansion for Ethereum's execution layer.
EIP-8037 and EIP-8038 — the proposals now under scrutiny — redesign how Ethereum charges for state-related operations. EIP-8037 creates a dedicated gas reserve for blockchain state creation; EIP-8038 adjusts costs for accessing and modifying existing state. The intent is to align gas charges more closely with actual resource consumption.
On October 13, ChainSecurity published an analysis identifying a class of attacks enabled by EIP-8037 and EIP-8038's revised gas accounting. The vulnerability centers on insufficient gas griefing — a technique where a malicious caller deliberately provides too little gas for a legitimate smart contract operation to complete.
The mechanism works as follows:
Under current Ethereum rules, when an operation runs out of gas, it fails and the calling contract typically halts. Some contracts are designed to catch failed sub-calls and continue execution — for instance, recording an unsuccessful attempt, removing a claim, or disabling retries. These patterns are safe today because gas depletion at a specific point in execution is predictable.
Under EIP-8037 and EIP-8038, gas may be refunded to the contract after an operation fails. This means a contract that previously would have halted can instead continue running with the refunded gas. An attacker can calibrate the supplied gas to cause a target contract to run out of gas at a precise execution point, triggering the failure-handling code path instead of a full revert.
The consequence: contracts that catch failed calls could permanently record an operation as unsuccessful — erasing a user's withdrawal claim, disabling a retry mechanism, or marking a batch operation as complete when it was not. According to ChainSecurity's analysis, these failures could become irreversible.
ChainSecurity identified 13 deployed smart contracts vulnerable to this pattern. The firm did not publicly name the affected contracts and is conducting responsible disclosure with their development teams. Pseudonymous researcher pcaversaccio described the risks as "serious" and called for hard fork reassessment on October 21, one day before the Ethereum Foundation announced the postponement.
ChainSecurity noted the changes "would not make all gas-related security checks unreliable," but that developers must review existing safeguards for compatibility with the new gas rules. The affected applications reportedly include DeFi protocols handling withdrawals, batched transaction processors, and contracts requiring automated failure recovery.
Glamsterdam's Sepolia testnet activation on October 6 at 13:53:36 UTC proceeded without reported issues. According to CoinDesk's analysis of over 25 consecutive post-fork blocks, each used between 52 million and 92 million gas (26% to 46% of the new 200 million capacity). During a six-minute voting period, Sepolia produced all 32 scheduled blocks, with 99.97% of eligible stake voting for finality.
However, Sepolia test conditions did not surface the EIP-8037/8038 vulnerability. The gas griefing risk requires interaction with specific contract patterns deployed on mainnet. ChainSecurity's October 13 disclosure — seven days after the Sepolia activation — identified the issue through manual analysis of mainnet contracts, not through testnet behavior.
The Hoodi testnet, Ethereum's final pre-mainnet testing environment, was scheduled for activation at epoch 132,352 (October 26, 17:42:48 UTC). The postponement removes three components: Amsterdam execution layer activation, Gloas consensus layer activation, and the gas-limit schedule increase to 200 million. Hoodi's existing chain history remains unaffected.
Node and validator operators on Hoodi must update both execution and consensus layer clients before October 26 to reflect the cancellation. Updated releases have been published for twelve client implementations: Grandine, Lighthouse, Lodestar, Nimbus, Prysm, Teku (consensus layer) and Besu, Ethrex, Erigon, go-ethereum, Nethermind, Reth (execution layer).
The postponement is particularly consequential for EIP-7732, which addresses a structural risk that has grown more acute during the delay. Ethereum's current block-production pipeline relies on MEV-Boost, an out-of-protocol relay system that mediates between validators (proposers) and specialized block builders.
The concentration data as of October 2026 underscores the urgency:
ePBS would eliminate the need for trusted relay intermediaries by settling proposer-builder coordination and payment natively at the protocol level. Each additional month that Glamsterdam is delayed extends the period during which Ethereum's block production depends on three builder operations and a handful of relay operators — infrastructure that exists entirely outside the protocol's security guarantees.
Prior to the ACDE #248 call, developers had been considering December 1, 8, or 9 for mainnet activation. The Hoodi delay removes all three dates from practical consideration.
The path to mainnet now requires: resolution of the EIP-8037/8038 compatibility issues (either through EIP modification, affected-contract migration, or a combination); rescheduling and successful completion of the Hoodi testnet activation; and a minimum observation period before mainnet commitment.
No developer or Ethereum Foundation representative has publicly stated a revised timeline. The postponement announcement explicitly noted that it "does not set or change any Ethereum mainnet activation date."
For context, Ethereum's Pectra upgrade — Glamsterdam's predecessor — experienced its own extended delay cycle. Originally targeted for April 2024, Pectra was repeatedly pushed back due to scope expansion and testnet failures on both Holesky and Sepolia, ultimately activating on mainnet May 7, 2025, approximately thirteen months late. Developers subsequently split the original Pectra scope into two phases, with the remaining features folded into Fusaka (activated mid-2026).
The Glamsterdam delay is, by comparison, early-stage — a testnet postponement rather than a mainnet slip. But the pattern of ambitious EIP bundling followed by security-driven delays has become a recurring feature of Ethereum's upgrade cadence.
ETH closed at $2,472 on October 21 and fell to $2,386 on October 22, a 3.5% decline on the day the postponement was announced. The token remained under pressure near $2,400 on October 23.
Attribution is imprecise. The decline coincided with $110.6 million in net outflows from U.S. spot Ethereum ETFs on October 22, with BlackRock's ETHA reportedly accounting for a significant share. An early Ethereum investor separately sold 3,000 ETH (approximately $7.2 million) during the same period. ETH had already declined approximately 11% over the course of October prior to the Glamsterdam announcement.
The broader staking environment shows 43.8 million ETH locked (35.9% of total supply) across approximately 863,000 active validators — down from 975,000 at year-start. The validator count decline reflects post-Pectra consolidation (validators can now hold above 32 ETH per entry, averaging ~50 ETH), not withdrawal pressure. Staked ETH actually grew from 36.3 million to 43.8 million during 2026, adding over 7 million ETH.
Ethereum has a documented history of upgrade timeline revisions:
| Upgrade | Original Target | Actual Mainnet | Delay | |---|---|---|---| | Constantinople | Nov 2018 | Feb 2019 | ~3 months (reentrancy bug) | | Pectra | Apr 2024 | May 2025 | ~13 months (scope split + testnet failures) | | Glamsterdam | Aug 2026 (initial), Dec 2026 (revised) | TBD | Ongoing |
The Constantinople precedent is instructive. In January 2019, ChainSecurity — the same firm behind the current disclosure — identified a reentrancy vulnerability in EIP-1283 just hours before the scheduled Constantinople hard fork. The upgrade was postponed and EIP-1283 was replaced with EIP-2200, which included additional safeguards. Mainnet activation occurred approximately one month later as part of the combined Constantinople/St. Petersburg upgrade.
The current situation differs in that ChainSecurity's disclosure came 13 days before the scheduled testnet activation (not mainnet), providing more runway for remediation. The scope of affected contracts (13 identified so far) is also more bounded than a systemic protocol vulnerability.
Glamsterdam represents a qualitative shift in Ethereum's upgrade ambition — ePBS, parallel execution, and repriced gas in a single fork. The delay is a direct consequence of that ambition. EIP-8037 and EIP-8038's gas changes interact with a deployed contract base that no testnet fully replicates, and ChainSecurity's finding that at least 13 live contracts face permanent failure modes under the new rules justifies the halt.
The practical question is remediation speed. If the EIP-8037/8038 issues require targeted modifications rather than a full redesign, a rescheduled Hoodi activation in November and mainnet in January 2027 remains plausible. If the changes expose deeper incompatibilities in Ethereum's gas model, the timeline extends further. Developers have not indicated which scenario they consider more likely.
What is clear is that Ethereum's block-production infrastructure — concentrated in three builder operations processing 92% of relay blocks — continues to operate outside protocol guarantees for every additional week that ePBS remains unshipped. The economic value secured by the network (43.8 million staked ETH, $51 billion in DeFi TVL) makes conservative upgrade sequencing rational. It also makes indefinite delay costly.