Solana validators restarted Mainnet Beta after consensus failure

Solana validator operators completed a coordinated restart of Mainnet Beta on October 1, 2022, restoring block production after a consensus failure had halted the network late on September 30. The official status update placed the completed restart at 07:00 UTC and warned that network operators and decentralized applications would continue restoring client services for several more hours.

The recovery was the consequential event attached to October 1: the halt began the prior calendar day, but the network returned and the first code-level remedy entered public review on October 1. That distinction keeps the archive date aligned with what actually changed during the UTC day.

What was established on October 1

The Solana Foundation’s later incident report timestamps the cluster halt at 22:41 UTC on September 30 and the return of an advancing root with 80% of stake weight online at 06:57 UTC on October 1. Subtracting those timestamps gives eight hours and 16 minutes between the recorded halt and that recovery threshold. That is a calculated interval, not a complete measure of user-facing downtime: voting problems were reported earlier, and wallets, remote-procedure-call providers and applications could remain impaired while client services were restored.

Contemporaneous reporting from Bloomberg confirmed the restart shortly after Solana’s status announcement. On October 1, however, the technical diagnosis was still developing. Reports centered on a validator operating a duplicate instance and producing conflicting blocks, but the full causal chain had not yet been formally published. It would therefore have been premature on October 1 to present every later postmortem detail as settled.

The public engineering record did show a concrete response. Solana Labs pull request 28172 was opened on October 1 to change how validators reset toward the heaviest fork after a duplicate slot. Its stated problem was that validators could fail to reset correctly when the heaviest bank’s slot matched their last voted slot. The proposal and a reproducing test were under review on October 1; the change had not yet been merged or broadly deployed.

Why the restart mattered

A general-purpose blockchain’s institutional value depends not only on throughput under normal conditions but also on continuous, predictable settlement. During the halt, the chain was not producing new confirmed state in the ordinary way. Applications could not treat the network as continuously available, and infrastructure operators had to coordinate around a shared restart point.

That made the episode both a software failure and a governance test. Stake-weighted validators, core developers, application operators and service providers had different operational roles, yet recovery required enough validators to adopt the restart plan for roots to advance. The event did not prove that one company unilaterally controlled all network state. It did show that, under a liveness failure, social coordination and common software procedures were essential to resume the ledger.

The market context was similarly narrow. Solana’s native SOL token traded continuously across multiple venues, but this reconstruction makes no price or percentage claim because no single official close exists and the cited incident records do not establish a causal market effect. What can be said from the October 1 record is that another extended interruption increased scrutiny of reliability at a time when Solana was competing for developers, users and institutional credibility.

What remained uncertain

On October 1, client restoration was still in progress, the proposed patch remained unmerged, and a complete postmortem was unavailable. The restart established renewed liveness; it did not by itself prove that the underlying consensus edge case had been removed or that another outage could not occur.

Later context

The Solana Foundation published its formal outage report on October 6. It attributed the trigger to a malfunctioning hot-spare configuration that allowed two validators with the same identity to produce different blocks, followed by an edge case in fork-selection logic that prevented consensus recovery. The linked fix was merged on October 3, and the Foundation later reported that 90% of stake weight had applied patched software by October 7. Those milestones clarify the October 1 event but were not yet knowable when the restart occurred.