On Wednesday morning, the Solana network briefly neared loss of transaction finality. A routing configuration issue at the custodial service Teraswitch caused 28.83% of staked SOL to go temporarily offline, and Solana halts block confirmation when approximately 33.34% of staked tokens are unreachable.
This means the network was just 4.51 percentage points away from completing finality. According to Marinade Finance’s calculations, this event reached approximately 86% of the “stall threshold.”
The outage was caused by a single point of routing failure.
Teraswitch later stated that the issue originated from a default route issued by its Miami site. This route lost its original metric and community attributes during propagation and was subsequently disseminated to European and Asia-Pacific nodes via an Amsterdam route reflector.
The local edge router misclassified it as a locally generated route, giving it higher priority than the normal exit; however, the data center core network deemed it invalid. As a result, 12 locations—including London, Amsterdam, Dublin, Frankfurt, Singapore, and Tokyo—temporarily lost valid forwarding paths, while North America remained unaffected.
The engineering team identified the issue within approximately 10 minutes, and service was restored at 04:16:15 UTC.
Nearly 30% of staked SOL went briefly offline
Marinade reported that approximately 90 validators were affected, collectively losing 333 SOL in rewards. Their analysis showed that the outage was highly concentrated on a single autonomous system, AS20326.
The system holds 118.89 million SOL, accounting for approximately 27.34% of the total network stake, with 94% of it going offline during the same period. This percentage exceeds the Solana Foundation’s delegated program limit of 25% for a single autonomous system.
Additionally, 59 validators with approximately 80.2 million SOL recovered within the same time window in Amsterdam, Frankfurt, and Tokyo, indicating that these nodes were primarily waiting for routing to reconverge rather than switching to alternative available paths. Solana’s second-largest validator, Helius, was offline for a full 33 minutes.
Concentration risk has been exposed again
Marinade also noted that an additional 14.1 million SOL were offline during the same period in托管 environments such as latitude.sh, Limestone, Butterfly Research, and Allnodes, a phenomenon that cannot be explained solely by the available data.
This means that measuring dispersion solely based on the custodian's口径 may underestimate the risk of validators becoming simultaneously offline during the same failure.
Marinade also disclosed concentration in its own allocation model: two-thirds of its staked tokens are concentrated in four autonomous systems, with AS395201 alone accounting for 36.94%. The protocol stated that it will reassess concentration limits by network and data center and begin disclosing whether validators have hot-swapping and automatic failover capabilities.
Additional information: The 333 SOL reward that was missed is expected to be covered by validator stakes after the current epoch ends. However, if the offline rate exceeds one-third at that time, transaction confirmations for all SOL holders will halt, and no staking mechanism exists to mitigate such systemic consequences. Solana’s last full network outage occurred in February 2024, when it took nearly five hours to restore operations.