Ethereum Postpones Glamsterdam Upgrade on Testnet Over Smart Contract Flaws

Ethereum core developers have officially halted the deployment of the next network hard fork to the Hoodi testnet, pulling the emergency brake after security researchers uncovered eight critical vulnerabilities in how proposed gas accounting rules interact with existing smart contracts. The decision, confirmed by the Ethereum Foundation on November 6, 2026, removes any remaining probability of a mainnet release before the end of the year and exposes deep tensions between base-layer scaling ambitions and smart contract backwards compatibility.

The target of the suspension is the Ethereum Glamsterdam upgrade, a multi-pronged overhaul combining the Amsterdam execution layer and the Gloas consensus layer. Glamsterdam was conceived as Ethereum’s most consequential protocol evolution since the September 2022 Merge, packaging native parallel transaction processing, enshrined proposer-builder separation (ePBS), and a massive block gas limit increase to 200 million units.

Yet during what was meant to be the final dress rehearsal on public testnets, smart contract security firm ChainSecurity discovered that two core components—Ethereum Improvement Proposals (EIP) 8037 and 8038—fundamentally destabilized longstanding developer assumptions regarding transaction execution, error handling, and gas stipends. Rather than risking billions of dollars locked across decentralized finance protocols on mainnet, core developers took the rare step of freezing the deployment pipeline just weeks after the code successfully activated on the Sepolia testnet.

The postponement marks a decisive turning point in Ethereum’s 2026 engineering cycle, shifting the developer community from aggressive throughput optimization back to defense-in-depth risk mitigation.

GLAMSTERDAM ESCALATION ROADMAP (2026)

[Jan - Feb]      Meta-EIP-7773 Formalized; Scope Defined
       │
[May 2]          Soldøgn Interop Devnet Reveals Early Client Friction
       │
[Aug 17-20]      Plataberget Devnet Launches; June/August Mainnet Targets Abandoned
       │
[Sep 14]         Devnet-11 Deployed to Address Execution Stalls
       │
[Oct 6]          Sepolia Testnet Fork Activates (Epoch 353,024)
       │
[Oct 13]         TURNING POINT: ChainSecurity Dossier Flags 8 "Gas Traps"
       │
[Late Oct]       GitHub 'glamsterdam-gas-invariants' Confirms Exploits
       │
[Nov 6]          EMERGENCY BRAKE: Ethereum Foundation Halts Hoodi Testnet Fork

Phase 1: Ambition and the 200-Million Gas Target (January – April 2026)

Ethereum's development roadmap entered 2026 riding high on the execution of two key milestones: the Pectra upgrade in mid-2025 and the Fusaka upgrade in December 2025. Fusaka introduced PeerDAS (Peer Data Availability Sampling), which dramatically expanded the data runway for Layer-2 rollups. However, it left foundational Layer-1 execution constraints untouched. Sequential transaction processing remained an unyielding bottleneck, third-party MEV-Boost relays introduced centralization vectors at the block-building layer, and unmetered state growth hovered near an unsustainable expansion rate of 120 gigabytes per year.

To solve these compounding structural problems, Ethereum Foundation researchers and client engineers coalesced around meta-EIP-7773, codenamed Glamsterdam. The fork followed Ethereum’s established naming protocol: Amsterdam, named after the host city of Devconnect, for the execution layer; and Gloas, named after the designated star, for the consensus layer.

The scope was audacious. The headlining proposals promised to transform Ethereum’s underlying runtime:

EIP-7732 (Enshrined Proposer-Builder Separation): Moves the block-construction market directly into the consensus layer, removing reliance on external relay software and dividing block creation into two distinct phases to protect validators from toxic maximal extractable value (MEV) extraction.
EIP-7928 (Block-Level Access Lists): Requires transactions within a block to pre-declare their memory and storage access footprints. By knowing ahead of time which transactions do not touch overlapping state, execution clients can process transactions concurrently across multiple CPU threads instead of processing them strictly in sequence.
EIP-8037 and EIP-8038 (State Creation and Access Repricing): A total overhaul of EVM economic metering. EIP-8037 instituted a hard cost-per-byte fee structure on newly created state, while EIP-8038 adjusted read/write storage opcodes (SLOAD and SSTORE) to reflect physical hardware access realities.

Engineers projected that parallel execution combined with state access lists would allow the base network to safely handle a block gas limit of 200 million units—more than triple the operational ceiling of roughly 60 million. Base-layer throughput was modeled to jump toward 10,000 transactions per second under optimal conditions, while simple transfers were expected to see fee reductions exceeding 70 percent via complementary proposals like EIP-2780 and EIP-7904.

The initial schedule targeted a mainnet deployment by June 2026. The development community believed that by bundling consensus-layer block pipelining with execution-layer repricing, Ethereum would answer criticisms regarding base-layer speed without compromising its node decentralization requirements.

Phase 2: Early Friction and Devnet Realities (May – August 2026)

The schedule began showing signs of strain during early cross-client interoperability trials. On May 2, 2026, developers concluded testing on Soldøgn, the first multi-client devnet engineered to trial ePBS alongside access lists. While consensus clients successfully negotiated the split payload structure dictated by EIP-7732, execution engines struggled under high transaction loads.

When block builders pushed access-list data into execution nodes, synchronization latency between consensus and execution engines caused block propagation windows to wobble dangerously. Individual nodes running Nethermind, Geth, and Besu frequently diverged on parallel execution state roots when handling conflicting transactions that were improperly indexed inside block-level access lists.

By early June, core developers conceded that the June launch window was impossible. The Ethereum Foundation revised its target to late August 2026.

The technical scope of the Ethereum Glamsterdam upgrade, however, continued to swell rather than contract. Auxiliary EIPs were introduced to fix auxiliary vulnerabilities:

EIP-7708: Implemented automated log emissions for native ETH transfers to eliminate blind spots in contract-to-contract transfers.
EIP-7997: Deployed a universal CREATE2 factory across all participating EVM networks to establish cross-chain contract deployment parity.
EIP-8070 (eth/72) and EIP-8189 (snap/2): Updated peer-to-peer networking sub-protocols to handle larger parallel state diff transfers between distributed nodes.

The escalating complexity triggered a domino effect across the testnet sequence. On August 17, 2026, the Ethereum Foundation deployed Plataberget, an ephemeral testnet built specifically to trial the combined stack. Three days later, on August 20, the hard fork executed.

Almost immediately, private devnet instances experienced state divergence stalls. Geth and Besu instances running at high capacity fell behind on state root recalculations whenever blocks exceeded 120 million gas. On September 14, client developers were forced to launch an emergency intermediary network, Devnet-11, solely to stabilize node consensus and fix client-level memory leaks.

With August long gone, core developers adjusted their mainnet timeline again, sliding the target into the fourth quarter of 2026. Yet despite the delays, client teams believed that the hardest obstacles were behind them. The protocol changes appeared mechanically sound, and the bugs were framed as client-level implementation issues rather than architectural design flaws.

Phase 3: The Sepolia Deployment and False Security (September – October 6, 2026)

On September 17, 2026, during All Core Developers Execution (ACDE) Call 198, core developers formally committed to the public testnet rollout calendar. The plan followed the protocol's traditional, sequential staging sequence: first Sepolia, the application-focused testnet, followed three weeks later by Hoodi, the primary validator testbed, and finally mainnet.

PUBLIC TESTNET PIPELINE (ORIGINAL TIMELINE)

[Sepolia Testnet] ───────► [Hoodi Testnet] ───────► [Ethereum Mainnet]
 Activated: Oct 6           Scheduled: Oct 27         Targeted: Q4 2026
 Status: SUCCESS            Status: HALTED            Status: POSTPONED

On September 28, the Ethereum Foundation published its official deployment announcement. Glamsterdam was slated to fork Sepolia at epoch 353,024, slot 11,296,768—arriving exactly at 13:53:36 UTC on October 6, 2026. Hoodi was tentatively scheduled for October 27, pending Sepolia’s network stability.

When the slot arrived on October 6, the activation executed with mechanical precision. Consensus clients—including Prysm 7.2.0 and Teku 26.9—seamlessly coordinated with updated execution clients. Block production under EIP-7732's enshrined builder separation performed as planned.

Within hours, validators on Sepolia pushed simulated blocks past the 150-million gas threshold, eventually hitting the full 200-million gas ceiling without dropping finality. Node operators celebrated. Ethereum’s base layer was executing transactions in parallel on a live public network, and gas throughput appeared to deliver on every paper promise.

Beneath the surface, however, the Sepolia success masked an existential testing blind spot. Sepolia is primarily populated by developer sandbox contracts, mock tokens, and synthetic test suites. The complex, legacy smart contracts that anchor mainnet—deeply nested lending vaults, automated market makers with reentrancy guards, proxy upgrade patterns, and multi-signature treasuries—do not maintain identical, live liquidity arrays on testnets. Sepolia proved that the clients could handle the new consensus and execution rules. It did not prove that the applications built over the past decade would survive them.

Phase 4: The Turning Point — ChainSecurity Exposes the Gas Traps (October 13 – October 28, 2026)

The sense of victory lasted exactly one week. On October 13, 2026, smart contract auditing firm ChainSecurity published a detailed vulnerability dossier that sent shockwaves through the Ethereum Foundation and client engineering teams.

The audit focused on the economic mechanics introduced by EIP-8037 (State Creation Gas Cost Increase) and EIP-8038 (State-Access Gas Cost Update). Under Ethereum’s current operational model, the gas cost of executing opcodes is static and predictable. Smart contract developers write Solidity and Vyper code under explicit assumptions about how much gas a call will consume, how out-of-gas (OOG) exceptions propagate, and what happens when an external subcall fails.

EIP-8037 and EIP-8038 fundamentally rewrote these economic constants to curb state bloat:

Operation / Opcode	Current Gas Cost	Glamsterdam Proposed Gas Cost	Percentage Increase
New Storage Slot Allocation (SSTORE)	22,100 gas	110,020 gas	+397.8%
Account Creation via ETH Transfer	25,000 gas	183,600 state gas	+634.4%
Contract Bytecode Deployment	200 gas / byte	1,530 gas / byte	+665.0%
Storage Slot Overwrite (Post-Read)	2,900 gas	Dynamic reservoir access (EIP-8038)	Variable / Tiered

ChainSecurity demonstrated that this aggressive repricing broke eight specific, widely deployed smart contract invariants—architectural patterns that production protocols rely on to prevent economic loss, maintain security boundaries, and ensure execution consistency.

EIP-8037/8038 GAS RESIDUAL FAILURE STATE

[Parent Contract Execution]
       │
       ▼
[External Subcall: Fixed Gas Allocation]
       │
       ├─► (Current Mainnet): Subcall hits expensive write ──► Fails out-of-gas ──► Parent Reverts Safely
       │
       └─► (Under Glamsterdam): Revised reservoir math alters execution frame
             │
             ▼
       [Subcall Partially Executes or Leaves Unexpected Gas Residuals]
             │
             ▼
       [Parent Assumes Clean Failure, Proceeds to Next Instruction]
             │
             ▼
       🚨 CRITICAL EXPLOIT: State desynchronization, trapped liquidity, or spoofed receipt.

The Eight Invariants at Risk

The technical analysis revealed eight failure modes that could expose live protocols to exploits or operational lockups:

1. Invalidation of the 2,300 Gas Stipend

For years, Solidity's .transfer() and .send() functions forwarded a hardcoded stipend of 2,300 gas to recipient addresses—just enough to emit an event or process basic logic, but purposefully insufficient to modify state or perform reentrant calls. Under EIP-8038, common read operations inside fallback functions exceeded this threshold, causing standard withdrawals from legacy contracts to revert universally. Conversely, where contracts used custom wrapper stipends designed to allow minimal logging, altered gas math opened unexpected reentrancy windows.

2. Revert-on-Depletion Assumptions

Multiple established DeFi protocols utilize a design pattern where an external call is expected to consume all remaining gas and trigger a revert if conditions are invalid. Developers assumed that an out-of-gas revert in a child frame would leave the parent frame with too little gas to continue, safely reverting the entire transaction. Glamsterdam’s revised dynamic gas reservoirs left unexpected gas reserves in the parent frame, allowing execution to proceed past logic gates that developers intended to be impenetrable.

3. Silent Try/Catch Failures in Yield Aggregators

Modern protocols frequently wrap external calls inside try/catch blocks. These blocks were engineered assuming that failures would stem from business-logic reverts (such as slippage limits). Under the massive cost increases of EIP-8037, subcalls began failing purely due to state allocation costs. The parent contract’s catch block would execute, misinterpreting an infrastructure gas exhaustion as an application-level rejection, leading to automated liquidation failures or incorrect bad-debt accounting.

4. Gas Traps in Batch-Processing and Multisig Queues

Corporate treasuries, DAO payout contracts, and decentralized payroll systems often batch multiple transfers into a single execution block. With account creation costs surging from 25,000 to 183,600 gas, transactions batching payments to previously empty addresses blew through individual transaction gas envelopes, permanently trapping scheduled multisig transactions in unexecutable queues.

5. Liquidation Vulnerabilities in Collateral Engines

Automated lending protocols (such as versions of Compound and Aave forks) enforce liquidations by calling liquidation engines that deploy dynamic storage slots to record seized collateral. Under a 110,020-gas allocation for new slots, liquidations near maximum gas limits failed to complete within acceptable priority fee boundaries, exposing lending pools to insolvencies during market drawdowns.

6. Broken Proxy Implementations (ERC-1967 and UUPS)

Certain proxy architectures utilize low-level assembly delegatecalls with gas checks based on the gasleft() opcode. The recalibrated gas decay curves under EIP-8038 altered the 63/64ths gas forwarding rule in unexpected edge cases, resulting in delegatecalls failing silently while returning a boolean flag indicating success to the caller contract.

7. Trapped Escrows and Fee Sweepers

Contracts designed to sweep residual token fees into newly minted vaults or treasury addresses became economically insolvent. The cost to deploy storage and initialize accounts exceeded the balance of the assets being swept, locking low-value tokens indefinitely inside intermediate routing contracts.

8. Asymmetric Fallback Exploits

In cross-contract interaction patterns where Contract A calls Contract B and expects an immediate callback, the disparity between memory reads and the updated storage writes allowed malicious actors to craft inputs that forced the callback into an out-of-gas state while the parent transaction finalized, tricking Contract A into releasing funds without receiving the corresponding asset transfer.

To confirm the severity of the findings, core developers and security researchers quietly launched a dedicated testing repository on GitHub, titled glamsterdam-gas-invariants. Within days, test suites verified what the developers feared: the exploits were not theoretical edge cases; they were reproducible across hundreds of verified mainnet smart contracts.

Phase 5: The Crisis Calls and The Emergency Brake (October 29 – November 6, 2026)

The revelation sparked urgent closed-door sessions among Ethereum core developers, client leads, and auditing firms. The community split into two opposing ideological and technical camps during All Core Developers Call 201 in late October:

The Scaling Realists: Several client developers argued that Ethereum could not indefinitely cater to outdated, legacy smart contract code at the expense of base-layer throughput. They noted that base-layer gas repricing had occurred before—most notably in EIP-150 (Tangerine Whistle) and EIP-1884 (Istanbul)—and argued that developers must adapt their contracts or migrate to modern proxy standards. They advocated for proceeding with the Hoodi testnet fork as scheduled on October 27, fixing edge cases on the fly.
The Application Security Conservatives: Security researchers, supported by decentralized finance builders and protocol leads, pushed back forcefully. They argued that unlike Istanbul, which bumped opcode costs by minor margins, EIP-8037 increased state creation costs by hundreds of percent. Immutable smart contracts holding billions in value cannot simply be patched; their rules are locked in bytecode forever. Deploying code that systematically alters execution invariants on immutable systems risked catastrophic exploit conditions across mainnet.

By the time ACDE Call 202 convened on November 5, consensus had shifted toward caution. Proceeding to Hoodi—a network intended to mimic mainnet conditions with real validator configurations—would have signaled to the market that the current upgrade specifications were locked.

On Friday, November 6, 2026, the Ethereum Foundation released its official verdict. In a statement published by core developer Marius van der Wijden, the Foundation announced that the Ethereum Glamsterdam upgrade was officially suspended from advancing to the Hoodi testnet.

"The Glamsterdam network upgrade will only move forward to the Hoodi testnet once the vulnerabilities identified within smart contract execution environments have been thoroughly analyzed, isolated, and mitigated," the Foundation stated. "While base-layer parallelization and ePBS remain our core scaling goals, preserving the cryptographic and functional integrity of deployed contracts on Ethereum mainnet is non-negotiable".

The announcement made it clear: no target block height or tentative date would be assigned to Hoodi until the specifications for EIP-8037 and EIP-8038 were fundamentally rewritten. The Q4 2026 mainnet window was dead.

Ecosystem and Market Fallout

The suspension sent immediate ripples across both financial and infrastructure layers of the Ethereum ecosystem.

In digital asset markets, the news removed a major technical catalyst that traders had anticipated for year-end performance. Ether traded flat to slightly lower following the announcement, hovering around $2,233. In crypto derivatives markets, short-term implied volatility rose sharply, while open interest in December call options shifted toward defensive put-spread positioning, signaling that institutional participants had priced out any 2026 throughput rally.

Across the decentralized finance developer landscape, however, the response was one of quiet relief. Major protocols—including lending protocols, automated market makers, and cross-chain bridge teams—immediately mobilized engineering resources to audit their code repositories against the test cases published in the glamsterdam-gas-invariants repository.

"Deploying an upgrade that introduces silent subcall failures would have been catastrophic," noted one senior smart contract auditor during an emergency developer roundtable following the announcement. "A testnet delay is an inconvenience. An unhandled gas trap that drains an automated market maker vault on mainnet is an existential crisis. The Ethereum Foundation made the only rational engineering call".

For node operators and client engineering teams, the freeze has created structural operational headaches. Teams behind Geth, Besu, Nethermind, Prysm, and Lighthouse had spent nearly a year aligning codebases for an integrated release train. Client teams must now decide whether to roll back testnet features on Sepolia, run dual-track development branches, or maintain the Sepolia fork in an artificial test state while core specifications are reworked.

The delay also complicates the operational economics of Layer-2 ecosystems. Networks such as Arbitrum, Optimism, and Base rely on Ethereum Layer-1 as their final settlement and data resolution court. Because many rollups execute transactions using modified versions of the Ethereum Virtual Machine, any divergence between L1 gas economics and L2 execution models threatens cross-chain state proof generation and bridge validity checks.

Technical Paths to Resolution

With the Hoodi testnet on hold, Ethereum core researchers are evaluating three divergent architectural paths forward to untangle the Glamsterdam deadlock:

THREE RECOVERY PATHWAYS UNDER EVALUATION

 Pathway A: Decouple & Ship
 ┌────────────────────────────────────────────────────────┐
 │ Ship EIP-7732 (ePBS) & EIP-7928 (BALs) in Q1 2027      │
 │ Drop EIP-8037/8038 Repricing to a Later Hard Fork      │
 └────────────────────────────────────────────────────────┘

 Pathway B: Soften the Gas Cliff
 ┌────────────────────────────────────────────────────────┐
 │ Keep Repricing, but Add State Subsidies / Rebates      │
 │ Introduce 2,300 Gas Stipend Forwarding Overhaul        │
 └────────────────────────────────────────────────────────┘

 Pathway C: The EOF Integration Pivot
 ┌────────────────────────────────────────────────────────┐
 │ Enforce New Gas Rules ONLY for EOF Bytecode Containers │
 │ Maintain Legacy EVM Opcode Economics for Old Contracts │
 └────────────────────────────────────────────────────────┘

1. Decoupling the Upgrade (The Fast Track)

The most direct path is to strip EIP-8037 and EIP-8038 from the scope of Glamsterdam entirely. Under this proposal, the hard fork would proceed strictly with consensus and parallel execution upgrades: Enshrined Proposer-Builder Separation (EIP-7732) and Block-Level Access Lists (EIP-7928).

Advantage: Client teams could deploy Hoodi within weeks and preserve a Q1 2027 mainnet timeline. Disadvantage: Without state creation repricing, core developers cannot safely raise the block gas limit to 200 million units. Raising capacity without curbing state bloat would accelerate disk storage growth toward unsustainable levels, threatening home stakers and validator decentralization.

2. Parameter Softening and Gas Stipend Subsidies

Rather than abandoning the repricing model, researchers could recalibrate the economic parameters. This would involve introducing dynamic gas subsidies for contracts executing fallback or receive functions, effectively restoring the purchasing power of the historic 2,300 gas stipend without abandoning the higher permanent state costs.

Advantage: Preserves the state bloat defenses of EIP-8037 while neutralizing the reentrancy and out-of-gas failure vectors. Disadvantage: Introduces complex state-transition special cases into client code, increasing client maintenance burden and code bloat.

3. The EVM Object Format (EOF) Enclosure

A third, more radical alternative involves linking the new gas pricing exclusively to EVM Object Format (EOF) contracts—a structured container format for EVM bytecode that has been debated for multiple fork cycles. Under this design, existing legacy bytecode would continue to operate under today's legacy gas costs, while only newly compiled EOF contracts would face the higher state creation fees and dynamic reservoir rules of EIP-8037/8038.

Advantage: Provides absolute backwards compatibility for every smart contract deployed on Ethereum since 2015. Disadvantage: Fractures the EVM execution environment into dual legacy and modern operational regimes, significantly increasing the complexity of client verification logic.

What to Watch Next

As Ethereum core developers regroup, the trajectory of the protocol’s roadmap hinges on several critical technical milestones over the coming months:

All Core Developers Consensus Calls (November – December 2026): Monitor bi-weekly ACDE calls for an official consensus decision on whether to decouple EIP-8037/8038 or redesign their state mechanics. A decision to decouple would point to an expedited testnet schedule; an architectural overhaul will guarantee an extended freeze.
The Status of the Sepolia Testnet: Because Glamsterdam remains live on Sepolia, watch whether client developers push a state rollback (a coordinated re-fork) or leave Sepolia on the active Glamsterdam spec while testing alternate patches. A testnet rollback would confirm that changes to the core specification will be extensive.
Rescheduling the Hoodi Fork: Hoodi testnet activation remains the indispensable gateway to mainnet. Until a specific epoch and slot are assigned for Hoodi, any mainnet projection is speculative. Historically, Ethereum requires a minimum of four to six weeks of flawless execution on Hoodi before scheduling mainnet parameters.
Knock-On Effects for Hegotá: Glamsterdam’s delay directly compresses the development timeline for Ethereum’s planned 2027 upgrade, Hegotá (combining the Heze consensus layer and Bogotá execution layer). Hegotá’s agenda—which includes native account abstraction refinements and censorship-resistance improvements—will likely be pushed into late 2027 or early 2028 as engineering resources remain concentrated on Glamsterdam.

The pause of the Ethereum Glamsterdam upgrade reinforces an enduring truth of base-layer blockchain development: scaling throughput is ultimately constrained not by cryptography or network hardware, but by the immutable history of the software that came before it. As developers work to untangle eight hidden gas traps, the priority remains clear: ensure Ethereum scales without breaking the foundation upon which its decentralized economy stands.