Ethereum core developers canceled the Glamsterdam hard fork activation on the Hoodi testnet during the ACDE #248 call on October 22, 2026. The upgrade had been scheduled for October 26 at 17:42:48 UTC (epoch 132,352). No replacement date has been set. The postponement followed an October 13 secur...
"A failure path that previously reverted can now complete… [turning] a retryable operation into a permanent failure." — ChainSecurity, Smart Contract Security Firm
Ethereum core developers canceled the Glamsterdam hard fork activation on the Hoodi testnet during the ACDE #248 call on October 22, 2026. The upgrade had been scheduled for October 26 at 17:42:48 UTC (epoch 132,352). No replacement date has been set.
The postponement followed an October 13 security report from ChainSecurity identifying 13 real-world smart contracts vulnerable to behavioral changes introduced by EIP-8037 and EIP-8038 — two proposals that restructure how Ethereum charges gas for state creation and state access. The vulnerability allows failure handlers inside contracts to complete execution where they previously would have reverted, enabling attackers to convert retryable operations into permanent failures.
ETH traded at $2,497.88 on October 22, up 2.5%, with a market capitalization of $304.76 billion. The market reaction was negligible. However, the delay jeopardizes Ethereum's December 2026 mainnet target — developers had been weighing December 1, 8, or 9 activation dates contingent on a clean Hoodi run — and introduces uncertainty into the timeline for the subsequent Hegotá upgrade.
Glamsterdam is Ethereum's next hard fork, succeeding Fusaka. It combines the Amsterdam execution-layer upgrade with the Gloas consensus-layer upgrade. The fork packages ten EIPs, anchored by two headline proposals:
Supporting these are EIP-8037 and EIP-8038, which reprice gas for state creation and state access respectively. The gas changes were designed to clear the path for raising Ethereum's block gas limit from the current 60 million toward a 200 million target — a precondition for meaningful L1 throughput gains.
Together, the upgrade's proponents had projected it could reduce MEV extraction by up to 70% (via ePBS), target throughput of 10,000 TPS (via BALs), and lower L1 fees by approximately 78.6% (via gas repricing under EIP-7904). Those projections now sit in limbo pending resolution of the security findings.
EIP-8037 introduces a cost-per-state-byte (CPSB) target of 1,530 gas per byte. Under current rules, contract creation costs a flat 32,000 gas. Under EIP-8037, it becomes 120 bytes × CPSB. Storage writes shift from a flat 20,000 gas to 64 bytes × CPSB. EIP-8038 applies similar restructuring to state access operations.
The intent is to align gas charges with actual resource consumption. The problem, identified by ChainSecurity on October 13, 2026, is that this restructuring changes how much gas remains available after a failed subcall. Under current rules, when an external call fails inside a contract, the gas consumed by the failed call leaves insufficient resources for the calling contract's failure-handling code to execute. The transaction reverts entirely — a safe default, because the state rolls back.
Under the proposed repricing, the gas arithmetic changes. Failure handlers that previously ran out of gas and reverted can now complete. A contract that was designed to revert on a failed subcall may instead record the failure as final, eliminating the user's ability to retry the operation.
ChainSecurity stated it had "already found 13 real-world examples" of affected contracts and was conducting responsible disclosure with the affected projects. The projects remain unnamed.
ChainSecurity's analysis identified three distinct mechanisms through which the vulnerability can be triggered:
1. State-Gas Reservoir Exploitation. EIP-8037 introduces a separate gas pool (the "state-gas reservoir") that travels with calls in full, while the forwarding limit applies to execution gas only. Solidity's gasleft() function reports only the execution balance, not the reservoir funding. Under current rules, approximately 15,000 execution gas would remain after a failed call — not enough for most failure handlers. Under Glamsterdam rules, approximately 22,100 execution gas would be available. According to ChainSecurity, "a handler that previously could not record a failure may now commit it."
2. Storage Refund Exploitation. Clearing a storage slot that was allocated earlier in the same transaction triggers what ChainSecurity described as "an immediate refill" under the new rules. This can restore up to 97,920 execution gas to the calling contract — more than enough for a failure handler to complete complex operations, including writing permanent failure states to storage.
3. Intermediary Gas Return. When a contract in the middle of a call chain reverts after a downstream target runs out of gas, the new rules credit outstanding execution-funded state-creation charges back to the intermediary's execution balance. This gas can then flow upstream to the original calling contract, enabling its failure handler to execute.
In all three cases, the practical impact is the same: an attacker can deliberately supply insufficient gas for an operation while leaving enough for the surrounding contract to record a permanent failure, destroying the recovery path.
The gas repricing exists for a reason. At the current 60 million gas limit, Ethereum's state growth is manageable. Geth's state database sits at approximately 390 GiB. But as the gas limit approaches 200 million — the target Glamsterdam was designed to support — state growth accelerates to an estimated 387 GiB per year without repricing.
Node operators face a degradation threshold at approximately 650 GiB. Without the repricing in EIP-8037, the runway from the current 390 GiB to 650 GiB is roughly 8 months at the 200 million gas limit. With EIP-8037's repricing targeting 120 GiB per year of state growth at 150 million gas, the runway extends considerably.
This creates the core tension: Ethereum needs the gas repricing to safely raise the gas limit, but the repricing introduces behavioral changes in existing contracts. The repricing cannot simply be dropped from Glamsterdam without also abandoning the 200 million gas target — and with it, much of the upgrade's throughput benefit.
The EIP-8037 Last Call deadline is November 1, 2026. Whether the EIP undergoes revision or the affected contracts are patched remains to be determined.
The delay also stalls deployment of EIP-7732, Ethereum's long-planned transition to enshrined proposer-builder separation.
Currently, 80-90% of Ethereum blocks are constructed by specialized builders and delivered to validators through off-chain relay networks. As of October 2026, block building is concentrated among three operations: Titan builds 53.0% of MEV-Boost blocks, Quasar handles 22.1%, and BuilderNet accounts for 16.1%. On the relay side, Ultra Sound relay processes 33.9% of payloads, Titan Relay handles 31.9%, and Aestus covers 16.2%.
This arrangement functions on a trust basis with no cryptographic guarantee. Under ePBS, builders become staked consensus-layer actors. A proposer includes a builder's signed commitment to an execution payload in the beacon block. The builder subsequently reveals the payload. A Payload Timeliness Committee attests to whether the payload was revealed on time. Payment flows through the protocol directly.
The relay infrastructure carries documented risks. Flashbots warned in August 2026 that the block building market is "running on fumes," with relay centralization intensifying. Censoring relays — those that filter transactions to comply with sanctions screening — have declined to 16.7% of MEV-boost payloads as of October 14, 2026, down from higher levels in prior years, while neutral relays now handle approximately 70% of daily block production.
ePBS does not depend on EIP-8037 technically, but it ships in the same fork. A delay to Glamsterdam delays ePBS regardless.
EIP-7928 introduces block-level access lists (BALs), allowing each block to declare upfront which accounts and contract storage it will read or modify. Client software can then preload the required state data and process transactions in parallel rather than sequentially.
The feature is designed to support the higher gas limits that the repricing EIPs enable. At 200 million gas, sequential transaction processing becomes a bottleneck. BALs address this by giving nodes advance knowledge of each block's state footprint.
Like ePBS, BALs are not directly affected by the EIP-8037 findings but cannot ship until the fork ships.
Before the cancellation, Glamsterdam had completed activation on the Sepolia testnet on October 6, 2026, with the 200 million gas limit tested successfully. The Hoodi activation was the second and final testnet stage before mainnet.
Following the postponement, the Ethereum Foundation directed node operators on Hoodi to either revert to older client releases or await updated versions with the upgrade removed. Updated client releases confirmed as of the announcement include Lodestar v1.50.1, Besu 26.10.1, Ethrex v30.0.1-rc.2, and Nethermind 2.2.1. Eight additional clients were still pending updates.
The Amsterdam execution-layer activation, Gloas consensus-layer activation, and the gas-limit schedule that would have raised Hoodi's gas limit to 200 million have all been removed from the testnet's configuration.
Prior to the cancellation, the All Core Devs call had planned to select from three candidate mainnet activation dates: December 1, 8, or 9, 2026. All three dates were contingent on a clean Hoodi testnet run.
With no replacement Hoodi date announced, the December mainnet window is now in jeopardy. Ethereum researcher pcaversaccio urged developers to "discuss the findings" and stated the discoveries warranted "a reassessment of this hard fork."
The delay also complicates the timeline for Hegotá (also referred to as Heka/Bogotá), the upgrade slated to follow Glamsterdam. According to a Yahoo Finance report, Hegotá was targeted for late 2026 as developers sought to accelerate the roadmap. A Glamsterdam slip into Q1 2027 would push Hegotá further into 2027.
The core developers face three paths forward: revise EIP-8037 to eliminate the behavioral changes in failure handlers, work with affected projects to patch their contracts before activation, or drop EIP-8037 and EIP-8038 from the fork entirely — forgoing the gas repricing and, by extension, the 200 million gas limit target.
The Glamsterdam delay exposes a structural problem in Ethereum's upgrade pipeline: the gas repricing needed for L1 scaling changes the behavioral assumptions embedded in live contracts. This is not a bug in the new code — it is an incompatibility between new gas arithmetic and existing contract logic that assumed specific failure modes.
The 13 affected contracts identified so far may represent the visible portion of a larger surface area. Every Ethereum contract that relies on gas-based failure handling to maintain recoverability is potentially exposed to the same class of vulnerability.
Core developers must now choose between patching the repricing EIPs, coordinating upgrades across affected contracts, or shipping Glamsterdam without its gas repricing component — each option carrying its own costs in terms of timeline, throughput, and backward compatibility. The EIP-8037 Last Call deadline of November 1 will serve as the next decision point.