Ethereum’s Kiln Testnet Completed the Merge—and Exposed a Client Fault

Ethereum’s Kiln testnet completed its Merge transition on March 15, 2022, moving from proof-of-work block production to a proof-of-stake network. The Ethereum Foundation described Kiln as the final dedicated Merge testnet expected before established public testnets were upgraded, making the exercise the closest public rehearsal yet for changing Ethereum’s live consensus system.

The transition finalized, but it was not clean. A Prysmatic Labs incident record dated March 15 documented a client encoding error that prevented Prysm validators in two tested client pairings from producing valid beacon blocks after the transition threshold was reached. That combination of progress and failure was the point of Kiln: demonstrate that the two-layer architecture could cross the boundary, then expose defects before they reached Ethereum mainnet.

What Kiln tested

Kiln began with an execution layer operating under proof of work and a separate Beacon Chain using proof of stake. At the configured terminal total difficulty, the proof-of-work chain stopped determining consensus and validators on the Beacon Chain assumed responsibility for proposing and attesting to blocks containing execution payloads.

The design required node operators to run an execution client and a consensus client together. The two components communicated through the Engine API. That pairing was operationally important because a fault in either side—or in the encoding passed between them—could interrupt block production even when the protocol transition itself proceeded.

The Ethereum Foundation said Kiln was intended for application developers, infrastructure providers, node operators and stakers to test full workflows. Most ordinary application interfaces were expected to remain stable, but proof-of-work-specific methods would not. Validators also needed an execution client to verify transactions and receive transaction priority fees after a mainnet Merge.

A successful transition exposed a real defect

Prysmatic Labs placed the transition threshold at approximately 15:00 UTC on March 15. Its postmortem said Prysm incorrectly converted the base-fee field in execution payloads between little-endian and big-endian formats. Execution clients then rejected the malformed payloads with an invalid-block-hash response.

The incident record estimated that the affected Prysm-Geth and Prysm-Nethermind validators accounted for approximately 15% to 20% of missed blocks. It reported no impact on attestation participation, while valid blocks carried more attestations to compensate for proposals that were not produced. Engineers identified the conversion problem, patched it, tested the change and released a new Kiln image.

Those percentages came from Prysmatic Labs’ testnet incident analysis, not an independently audited mainnet dataset. They describe the observed Kiln configuration around the transition and should not be generalized into a forecast of mainnet failure rates.

Client diversity helped the test network continue finalizing while Prysm proposers were impaired. That result did not prove every client pairing was production-ready. It demonstrated both the resilience benefit of multiple implementations and the coordination cost of requiring independently built clients to interpret the same payloads identically.

What March 15 did not settle

Kiln did not change Ethereum mainnet on March 15. Miners continued producing mainnet blocks, users did not need to move funds, and Beacon Chain withdrawals were not enabled by the rehearsal. The Ethereum Foundation had also set no date for the mainnet transition. Its stated sequence still required client fixes, transitions on established public testnets and stabilization before developers could set the mainnet activation threshold.

The defensible event-day conclusion was narrower than “Ethereum completed the Merge.” Ethereum completed a proof-of-stake Merge on Kiln and found an implementation defect under realistic cross-client conditions. That was a consequential engineering milestone, not a production launch.

No ether return, trading-volume change or broader crypto-market move is attributed to Kiln here. Continuous trading and overlapping macroeconomic news make an event-day causal market claim unsupported by the reviewed records.

Later clarification

On March 23, 2022, the Ethereum Foundation confirmed that Kiln had become a functioning proof-of-stake testnet while noting Prysm’s invalid blocks and execution-client synchronization errors. That later technical summary supports the March 15 record without implying that the eventual mainnet outcome was knowable on the test date.