Ethereum’s Soldøgn tests exposed the conditions behind a 200M gas target

Ethereum core contributors reported on May 2, 2026 that a multi-client test network had run the Glamsterdam upgrade’s external-builder pipeline end to end, while two disputed questions remained around the proposed enshrined proposer-builder separation system.

The Ethereum Foundation’s Soldøgn recap described the result as meaningful progress toward a credible post-Glamsterdam gas-limit floor of 200 million. It did not describe that figure as an activated mainnet limit, a final protocol parameter or a guaranteed outcome. Production readiness still depended on specification decisions, broader testing and approval through Ethereum’s public AllCoreDevs process.

What the interop tests established

Just over 100 core contributors gathered in Longyearbyen, Svalbard, for work conducted from April 27 through May 1, 2026. According to the May 2 report, teams began by attempting a development network containing four execution-layer clients and four consensus-layer clients. Initial problems delayed that target, but a four-by-three configuration became stable enough for stress testing on April 28.

Testing then followed a repeated cycle of stressing the network, finding cross-client inconsistencies and implementing fixes. One new test suite exposed a gap shared by every client implementation in the handling of execution-request invalidation of beacon requests. By May 1, nearly all participating clients were running together on a second Glamsterdam development network, and the external-builder pipeline had been exercised end to end.

These were development-environment results. They demonstrated interoperability under the workshop’s test conditions, not security or reliability across Ethereum mainnet’s full range of network conditions, validator behavior and adversarial activity.

Why ePBS mattered to scaling

Enshrined proposer-builder separation, specified through EIP-7732, would move the relationship between block proposers and specialized block builders into Ethereum’s consensus protocol. The design separates the consensus block from the execution payload and introduces explicit deadlines for bids, payload disclosure and attestations.

That sequencing could give clients more time to execute and validate payloads, providing part of the technical headroom for larger blocks. The broader Soldøgn plan combined ePBS with block-level access lists, which expose a block’s state-access pattern for parallel processing, and EIP-8037, which raises state-creation costs to constrain database growth as capacity rises.

The reported 200 million figure was therefore a combined engineering judgment. It depended on consensus timing, execution-client benchmarks and repricing controls working together. It was not evidence that throughput would increase proportionally in every workload or that application fees would fall by a specified amount.

Two questions remained open

The May 2 report identified two ePBS matters as genuinely contested. Developers had not settled whether a request signature should commit to the specific builder receiving it. They also had not resolved how a design allowing builders to participate with a 1 ETH stake could resist peer-to-peer Sybil attacks intended to damage network liveness.

Those questions were material because ePBS would place builder registration, payments and payload availability closer to Ethereum’s consensus rules. A development network can show that compliant software communicates successfully while leaving open how the system behaves when participants create many identities, withhold messages or exploit ambiguous commitments.

What the 200M number did—and did not—mean

Soldøgn’s organizers had initially described their goal only as finding an unspecified “X00M” conservative floor. The May 2 recap supplied 200 million as the workshop’s credible post-upgrade target after benchmarking and interoperability work.

The Ethereum Foundation nevertheless said final values would be decided through AllCoreDevs calls. No activation date was announced on May 2, and the report said draft changes still needed to become merged, production-ready code with finalized test coverage.

Later context

Coinburn’s May 11, 2026 archive record separately covers the broader Glamsterdam devnet milestone. This reconstruction preserves the distinct May 2 record: the successful interoperability test was accompanied by unresolved security and commitment questions that still constrained the path from workshop result to mainnet change.