Ethereum currently lists October 6, 2026 as the next major Glamsterdam milestone: the upgrade of Sepolia.

The official roadmap still targets mainnet for Q4, with no confirmed activation date.

Before that happens, developers want the new proposer-builder architecture exposed to a much less controlled environment than the early multi-client devnets.

That exposure has produced an interesting testnet-specific problem.

Glamsterdam is moving the builder relationship into Ethereum itself

Enshrined Proposer-Builder Separation, or ePBS, is one of the headline changes in EIP-7732.

A proposer handles the consensus block while a specialized builder assembles the execution payload and submits a bid.

Ethereum already has specialized block builders today, but much of the exchange between builders and validators depends on off-protocol software such as MEV-Boost relays.

Glamsterdam makes key parts of that relationship enforceable by the protocol.

The execution payload is deliberately separated from the initial selection

The architecture does more than formalize payments.

A proposer can select a builder commitment before the entire execution payload has propagated.

A Payload Timeliness Committee separately helps determine whether that payload arrived when expected.

Ethereum's roadmap says this dual-deadline structure expands the useful propagation window from roughly two seconds to around nine.

That extra time is a major part of the scaling rationale for ePBS.

But economic commitments behave differently when the currency is test ETH

On mainnet, irrational builder behavior has a real cost.

Deposits, bids and other economic consequences are denominated in ETH that people actually value.

Sepolia's ETH exists so developers can test systems without paying real money.

Faucets and test allocations are a feature of the network.

They also weaken any defense whose practical strength depends partly on making malicious behavior expensive.

The attack is almost comically straightforward

A malicious operator can create a builder identity, obtain test ETH and submit a bid far above what a normal builder would rationally offer.

The proposer selects the winning bid.

The builder then fails to reveal the promised execution payload.

Instead of competing to build the most valuable block, the attacker competes to win the right to withhold one.

Disposable identities make reputation much less useful

An ordinary production builder wants to remain usable after the current slot.

A testnet attacker does not necessarily care.

During an Ethereum core-developer discussion, consensus developer Potuz described the possibility of rotating a large number of builders, submitting very high bids and repeatedly failing to produce payloads.

If one builder identity becomes distrusted, another can take its place.

That turns identity-level blocking into an incomplete defense unless clients can respond quickly enough to the broader behavior.

This is not a way to steal mainnet ETH

The scope needs to remain clear.

The scenario under discussion concerns Sepolia testing.

It does not give an attacker access to mainnet balances.

It also does not mean a builder can arbitrarily convince honest clients to accept invalid execution.

The immediate risk is degraded block production and a noisier infrastructure test.

A missing payload is already a state the protocol has to understand

ePBS cannot be designed on the assumption that every selected builder behaves honestly.

The specification includes handling for cases where the execution payload is withheld or fails to arrive on time.

Consensus must remain coherent even when the second half of the builder commitment fails.

The Sepolia concern is therefore less about discovering that payload withholding exists and more about how cheaply an adversary can force the network into that degraded path again and again.

Fallback behavior is only one half of the answer

Client implementations already have mechanisms for surviving missing payloads and developers have been exercising pathological consensus conditions throughout Glamsterdam's devnet cycle.

The Ethereum Foundation explicitly said in August that non-finality testing would be used to explore these unusual failure modes.

The new discussion focuses on a different operational question: how fast should software stop accepting bids from a builder that has repeatedly failed to deliver?

Surviving the failure is useful. Avoiding the same participant immediately causing it again is better.

A test network cannot perfectly reproduce mainnet economics

That is not a defect that can simply be eliminated.

Test ETH has to be accessible or developers would need to purchase a scarce asset merely to test contracts and infrastructure.

The moment a protocol mechanism uses economic cost as part of its security assumptions, however, a free testing currency can distort the experiment.

ePBS makes that distortion unusually visible because builders become explicit economic actors inside the consensus protocol.

This is exactly why Sepolia exists

Finding the scenario before mainnet is not evidence that the testing process failed.

It is evidence that moving from controlled devnets into public long-lived testnets changes what kinds of adversaries appear.

Platåberget already gives the public a place to deposit validators and builders and run post-Glamsterdam infrastructure.

Sepolia adds a much larger surrounding ecosystem of clients, applications and operational tooling.

A wider test environment is supposed to expose assumptions that smaller rehearsals miss.

The builder change is already a large operational migration

Staking pools, validators, block-building infrastructure and monitoring systems all need to understand the new ePBS flow.

Execution and consensus clients both require Glamsterdam-compatible releases.

Ethereum.org explicitly warns node operators that they must upgrade before activation in order to remain synchronized.

Builder-withholding behavior therefore arrives on top of an already significant change to the block-production pipeline.

And Glamsterdam changes much more than the builder market

The fork also includes Block-Level Access Lists, state-related gas repricing and larger limits for deployed contracts and initcode.

Some of those changes reach well beyond validators.

The Ethereum Foundation has warned that wallets, indexers and gas estimators relying on a hard-coded maximum gas limit need to update their assumptions.

Sepolia will consequently be testing application compatibility and execution-layer behavior at the same time that ePBS is being tested under more realistic public conditions.

The Sepolia release window is unusually tight

Client teams are currently targeting September 29 for Sepolia-ready releases, according to the developer schedule discussed this week.

That leaves seven days before the October 6 fork.

Ethereum upgrade processes normally aim for a longer period for review and bug-bounty testing.

Developers accepted the compressed Sepolia window in part because a testnet is easier to recover than mainnet if something fails.

It still puts more pressure on the days immediately following the client releases.

Hoodi is still supposed to come before mainnet

Developers have discussed October 27 as a tentative Hoodi date.

That timing remains conditional on what happens on Sepolia.

The official Ethereum roadmap does not yet present a confirmed Glamsterdam mainnet day.

Q4 remains the target rather than a guarantee.

A hostile testnet economy may actually be useful

The easiest ePBS test would have every builder behave rationally and reveal every payload.

It would also teach considerably less.

A public network where somebody can acquire worthless ETH and spend irrational amounts of it creates adversarial behavior that is hard to reproduce naturally in a polite interoperability lab.

That can make Sepolia unstable or annoying.

It also forces client teams to confront a builder whose objective is disruption rather than profit.

ePBS ultimately has to tolerate builders that do not behave economically

The long-term goal is to replace important off-protocol trust relationships with behavior Ethereum can reason about directly.

That means defining what happens when a builder bids, pays, reveals, arrives late or withholds entirely.

Mainnet adds strong economic pressure around those rules.

Sepolia removes much of that pressure.

The resulting environment can be harsher in one very specific way: participants can behave financially absurdly at almost no cost.

October 6 will therefore test incentives as much as software

Glamsterdam is often discussed in terms of throughput: longer payload propagation windows, parallelization and a path toward much higher L1 capacity.

Sepolia is about to test something less visible underneath those goals.

Ethereum is putting an actual proposer-builder market into consensus.

Markets are made from protocol rules, but they also depend on participants caring about what they gain and lose.

Test ETH deliberately weakens that second condition.

If ePBS and its clients remain usable while disposable builders throw valueless money at auctions and then refuse to do the job they won, Sepolia will have tested a much more interesting system than one where everybody politely follows the happy path.