Cardano recovered after roughly half its nodes restarted

Cardano’s mainnet recovered without a coordinated restart after a software fault caused roughly half of its block-producing and relay nodes to disconnect or restart on January 22, 2023. The disruption briefly interrupted block production but did not leave the network offline for an extended period.

At approximately 00:09 UTC, between blocks 8,300,569 and 8,300,570, operators began reporting failures in cardano-node version 1.35.4. A GitHub issue opened on January 22 preserved logs from one stake-pool operator whose public nodes with incoming connections shut down after a fatal Data.Map.balanceR exception. The operator said nodes without incoming connections were unaffected in that deployment.

Input Output Global’s contemporaneous message to stake-pool operators, reproduced by reports published on January 23, estimated that approximately 50% of Cardano nodes disconnected and restarted. It said both relay nodes and block-producing nodes were affected, while edge nodes appeared unaffected. Some nodes lost peer connections; others threw an exception and restarted.

A short disruption with a broad footprint

The most important event-day fact was the recovery path. A large share of a single blockchain’s operational infrastructure encountered the same anomaly, yet affected nodes generally returned without operators coordinating a network-wide reboot. Block production resumed after a brief degradation.

The approximately 50% figure was an IOG estimate, not a complete independent census of every Cardano node. The GitHub record directly documents failures on one operator’s machines and notes confirmation from other operators; it does not by itself measure the global share. The broader estimate is supported by IOG’s operator communication and was later affirmed in the project’s formal incident report.

That distinction matters. A network can remain available while still suffering a serious infrastructure incident. Relay nodes propagate data among peers, and block-producing nodes add blocks under Cardano’s proof-of-stake design. A correlated failure across both categories exposed implementation concentration: many independent operators were running closely related node software, so a shared defect could affect them simultaneously.

At the same time, the automatic return of affected nodes showed that Cardano’s consensus and node-recovery design could absorb this particular transient failure. That outcome was evidence of resilience under one real incident, not proof that the system could withstand every bug, attack or wider loss of connectivity.

What was known on January 22

The root cause was not established publicly on January 22. IOG described the trigger as a transient anomaly and said engineers were investigating. Claims that a particular transaction, attack or software component had definitively caused the event went beyond the contemporaneous record.

There was also no event-day evidence in the cited records of a prolonged chain halt or a coordinated validator restart. The defensible conclusion was narrower: block production was briefly impaired, many relay and producer nodes restarted or disconnected, and the network resumed operation on its own.

Later technical context

A formal incident report published on April 17, 2023 put the block-production pause at approximately two minutes and classified the incident as critical because it had potential to affect network integrity. The report attributed the fault to a complex data-structure bug involving the insertion and deletion order of multi-asset tokens in an internal ledger record.

That diagnosis was not available for event-day framing. The later report also said a fix was tested and shipped in cardano-node 1.35.5 on January 27, 2023. Those later findings confirm the event and explain it, but they should not be read backward as facts known at 00:09 UTC on January 22.