Ethereum finality faltered for roughly 25 minutes as participation plunged

Ethereum’s proof-of-stake network temporarily stopped finalizing blocks on May 11, 2023, while continuing to produce blocks and process transactions. Historical consensus-layer records place the disturbance between approximately 20:06 and 20:32 UTC. The interruption lasted roughly 25 minutes and exposed a distinction that matters to exchanges, bridges and other institutions: a blockchain can remain live while its strongest settlement guarantee is temporarily unavailable.

Finality returned without manual intervention. At the end of May 11, Ethereum developers had not established the root cause. The verified event-day conclusion was therefore limited: validator participation had fallen sharply, blocks were still being proposed, and the chain eventually resumed finalizing.

What the network record showed

Beaconcha.in’s record for epoch 200,551, which began at 20:06:47 UTC on May 11, shows voting participation of 88.4%. Of the epoch’s 32 slots, 27 were proposed, four were missed and one was orphaned. The explorer recorded 566,483 validators for that epoch.

The deterioration continued. A subsequent Prysm client-team reconstruction placed participation at 69% in epoch 200,552 and 40% in epoch 200,553. It reported that 18 of the 32 slots in epoch 200,553 were missed. Epoch 200,555 began at 20:32:23 UTC, when the network started recovering without intervention. Those measurements describe the consensus layer rather than an exchange’s ETH market, and they should not be interpreted as price, volume or liquidity data.

Contemporaneous reports published on May 11 said finality had returned within half an hour but that developers were still investigating. That uncertainty is important. Claims about a specific software defect, trigger or affected-client share were not established by the close of the event date.

Why losing finality mattered

Ethereum validators vote on blocks through attestations. When enough stake supports the required checkpoints, the protocol finalizes part of the chain, giving applications a strong assurance that those blocks will not be reorganized under normal consensus assumptions.

During the May 11 disruption, recently produced blocks lacked that finalized status for longer than usual. Users could still submit transactions and blocks continued appearing, so describing the episode as a complete Ethereum shutdown would be inaccurate. Nevertheless, applications that depended on finalized Ethereum state had reason to wait. An exchange crediting deposits, a bridge accepting messages or an institution treating an on-chain transfer as settled could not obtain the normal finality signal during the affected window.

The episode also made client diversity operational rather than theoretical. Validators were running multiple independently developed consensus clients, and contemporaneous observers argued that unaffected implementations helped the network continue operating. On May 11, however, that was an informed interpretation, not yet a completed root-cause finding.

What remained unknown on May 11

Developers had not determined whether the disturbance originated from a defect, an unusual network condition or an interaction among clients. There was also no verified evidence in the reviewed records that the incident caused a particular ETH price move. This reconstruction therefore makes no return, volatility or trading-volume claim and does not infer market causation from the timing alone.

A separate finality disruption occurred on May 12, 2023. It is not part of the May 11 event record and should not be used to enlarge the duration or impact assigned to this date.

Later technical context

A Prysm-team post-mortem published on May 18 attributed the disruptions to expensive processing of valid but old attestations, which exhausted resources in some consensus clients. It estimated that approximately 47 blocks were missed during the first incident for reasons attributable to the identified cause. Prysm’s v4.0.4 release, published on May 15, included resilience fixes connected to the mainnet incidents. These later findings clarify the May 11 record; they were not available when finality first returned.