Solana's network neared a complete transaction halt Wednesday morning when a misconfigured default route at hosting provider Teraswitch took 28.83 percent of its staked SOL offline. The incident brought the blockchain within 4.51 percentage points of its 33.34 percent finality threshold.
Teraswitch later published a technical post-mortem detailing the failure's origin. The company uses an internal default route to signal that an edge router can reach the internet. Each site within the Teraswitch infrastructure normally prioritizes routes originated by its own local routers.
A route from Teraswitch's Miami headquarters was propagated across the network without its metric or community attributes. That omission bypassed standard routing protocols. A route reflector in Amsterdam then distributed the malformed route to Europe and Asia-Pacific.
Routers in those regions interpreted the incomplete route as locally originated, overriding their preference for valid local paths. They forwarded that invalid route to the data center core, which rejected it—leaving 12 Teraswitch sites without any valid internet routes.
Affected data centers included London, Amsterdam, Dublin, Frankfurt, Singapore and Tokyo. Engineers identified the routing fault within 10 minutes. Solana's service was fully restored at 04:16:15 UTC.
Staking protocol Marinade Finance conducted a post-incident analysis of on-chain data. Its findings showed the disruption concentrated within a single autonomous system, AS20326, which holds 118,890,767 SOL—over 27 percent of total staked supply. Some 94 percent of stake within AS20326 went offline within the same minutes as the Teraswitch failure.
That concentration directly violates the Solana Foundation's delegation program guidelines, which cap exposure per autonomous system at 25 percent to limit single points of failure.
Marinade measured 74 affected operators. Of those, 59 validators collectively holding 80.2 million SOL waited for routing reconvergence rather than migrating to alternative infrastructure. Helius, the network's second-largest validator, remained inactive for the full 33 minutes of the outage.
Only three operators resumed activity without disruption: Laine, Cogent Crypto and Lion3d. Laine and Cogent Crypto are both managed by Sol Strategies, giving them operational redundancy the majority of validators lacked.
Validators affected by the downtime collectively lost 333 SOL in rewards—approximately $25,309 at $75.95 per SOL. The Solana Foundation's validator bonds are set to cover those losses at the close of the current epoch.
The incident exposed a gap in that coverage: no compensation mechanism exists for a full finality halt. Had offline stake exceeded one third of the total, no transactions would have finalized across the network. The bond structure does not address that scenario.
Marinade said it will revise its own concentration limits by network and data center following the event. The firm also committed to disclosing whether validators operate with hot swap and automatic failover capabilities—information that is not currently verifiable externally.
Marinade's own allocation model showed concentration risk: four autonomous systems hold two thirds of the stake it distributes, with AS395201 alone accounting for 36.94 percent of Marinade's allocated stake.
