Solana came within 4.51 percentage points of losing transaction finality on Wednesday morning after a single misconfigured route at hosting provider Teraswitch knocked 28.83% of staked SOL offline. The chain stops finalizing at 33.34%, which means the network got about 86% of the way to a halt.
Teraswitch traced the incident to a default route propagated from Miami
Teraswitch published a writeup describing how the failure unfolded. The company said it uses a default route internally to signal that an edge router can reach the internet, and each site normally prefers the route originated by its own routers.
According to the company, a default route from its Miami site was propagated after its metric and communities were stripped. A route reflector in Amsterdam then pushed that route into Europe and Asia-Pacific.
Edge routers in those regions treated the route as if it had been originated locally and preferred it over the valid one. They then passed it to the data center core, which rejected it as invalid. That left 12 sites in London, Amsterdam, Dublin, Frankfurt, Singapore and Tokyo without any valid path to forward traffic. North America was unaffected.
Engineers identified the fault within 10 minutes, and service returned at 04:16:15 UTC.
Marinade counted 333 SOL in missed rewards across 90 validators
Staking protocol Marinade Finance, which analyzed the incident afterward, said the event barely registered in broader discussion even though Solana had moved most of the way toward a halt. It reported that 28.83% of staked SOL went delinquent, while finality stops at 33.34%.
Marinade added up the rewards lost across the 90 affected validators and put the total at 333 SOL.
One autonomous system carried more than a quarter of network stake
Marinade found that the outage was concentrated in a single autonomous system. AS20326 carries 118,890,767 SOL, more than a quarter of all stake on the network, and 94% of that amount went offline during the same minutes.
That already exceeds the intended ceiling. The Solana Foundation delegation program caps any one autonomous system at 25% of network stake, while AS20326 stands at 27.34%.
Roughly 80.2 million SOL waited for routing to reconverge
Marinade said 59 validators holding 80.2 million SOL came back within the same narrow window in Amsterdam, Frankfurt and Tokyo. Those validators waited for routing to reconverge rather than switching elsewhere.
Helius, the second-largest validator on Solana, remained down for the full 33 minutes. Of the 74 operators Marinade said it could measure, only three “came back clean”: Laine, Cogent Crypto and Lion3d. Laine and Cogent Crypto are both run by Sol Strategies.
The outage pattern extended beyond one hosting provider
Marinade also identified another 14.1 million SOL that went offline in the same minutes across latitude.sh, Limestone, Butterfly Research and Allnodes. The firm said it could not explain those cases from the available data, and argued that measuring stake concentration by hosting provider can understate how much fails together.
Marinade applied the same concentration analysis to itself
Marinade said that four autonomous systems hold two-thirds of the stake distributed by its own allocation model, with one of them, AS395201, accounting for 36.94%.
The firm wrote on X, 「Nobody should be comfortable with that, us included,」 and said it would review its concentration limits by network and by data center. It also said it would begin publishing whether a validator runs hot swap and automatic failover, information that cannot currently be determined from outside.
If delinquency had crossed one-third, finality would have stopped network-wide
Marinade said the 333 SOL in missed rewards will be covered by validator bonds at the end of the epoch.
Had delinquency crossed one-third, nothing would have finalized for any SOL holder anywhere, and no bond covers that scenario. According to Decrypt, the last time Solana halted outright was in February 2024, and it took close to five hours to restart.