EOS mainnet halted after a deferred-transaction fault broke consensus

The EOS mainnet stopped producing blocks at 09:56 UTC on June 16, 2018, less than two days after token voting activated the newly launched network. A joint status statement from the top 21 block producers and standby nodes said the producers detected a break in consensus and halted together, stopping normal transaction processing while developers and operators investigated.

The interruption mattered because EOS had entered operation only on June 14 after the 150 million-token voting threshold—15% of the supply—was reached. The launch was an unusually public test of delegated proof of stake: token holders elected a limited active set of producers, and those operators were responsible for keeping a high-profile smart-contract network in agreement.

At 10:01 UTC on June 16, producers and standby operators joined a conference call. At 10:57 UTC, standby producers were told to stop nodes and preserve diagnostic information. The joint statement said incoming network connections were not being accepted during the response.

A deferred-transaction fault broke consensus

At 13:28 UTC, the operators attributed the incident to the handling of deferred transactions—transactions scheduled by smart contracts for later execution. The EOSIO repository’s contemporaneous pull request 4158 provides the code-level record of the fix. That primary record is stronger evidence than early headlines that described only a generic freeze, but the surviving public material does not establish which user action first triggered the faulty path.

The operators said the software paused when producers detected a consensus break, a safeguard intended to prevent producers from extending incompatible histories into a hard fork. That framing was a contemporaneous claim from the network’s operators. What can be independently verified from the retained timeline is that block production stopped, a patch was prepared, and the mainnet branch was advanced to version 1.0.5.

The recovery plan required producer nodes to upgrade, affected producers to resynchronize through block 1,027,926 with production paused, and producers to reconnect before public access reopened. Block.one issued the 1.0.5 tag at 14:13 UTC. The operators reported the network live again at 14:48 UTC. Calculated from the stated halt and restart times, the interruption lasted 4 hours 52 minutes.

A large network met an early operational test

The outage did not prove that delegated proof of stake could not work. It did show how EOS’s safety and recovery depended on fast coordination among a small elected producer set and the developer of the underlying software. That was both a practical advantage and a governance concern: the participants could organize a same-day repair, but ordinary users had no independent route to transact while the producers had stopped and closed access.

Market scale made the test consequential. CoinMarketCap’s historical snapshot for June 17—the first retained weekly snapshot after the incident—ranked EOS fifth by reported market capitalization. It displayed EOS at $10.44, with a reported market capitalization of $9.359 billion, circulating supply of 896,149,492 EOS, and $432.121 million in 24-hour volume. The snapshot showed a 2.15% 24-hour decline.

Those figures are an aggregated next-day snapshot, not an event-time quote or a consolidated regulated close. CoinMarketCap does not identify an exact June 16 outage-window price on that page, and cryptocurrency traded continuously across venues. The 24-hour decline therefore cannot be attributed to the halt from this record alone.

What remained uncertain on June 16

By 14:48 UTC, the immediate operational question was answered: producers reported block production restored after installing the patch. More difficult questions remained unresolved on June 16, including whether the same deferred-transaction path could recur, how independently producers verified the fix, and whether the coordination model concentrated too much emergency authority.

Later context

An EOSIO issue opened on October 22, 2018 documented additional problems involving replacement of deferred transactions. That later record should not be treated as proof that those separately described defects caused the June 16 halt. The event-day evidence supports a narrower conclusion: a deferred-transaction handling fault interrupted EOS consensus, the elected producers coordinated a controlled halt, and version 1.0.5 returned the network to service in under five hours.