A misconfigured default route at hosting provider Teraswitch knocked nearly 29% of staked Solana offline on Wednesday. The network came within about 20 million tokens of the one-third stake threshold where Solana stops finalizing transactions.
That is 86% of the way to a full finality loss, per Marinade Finance, the liquid staking protocol whose post-mortem on X underpins both the CoinDesk and Decrypt coverage. Solana needs more than two-thirds of staked tokens online to finalize blocks. Push one-third offline and the chain stops confirming transactions for every SOL holder.
The bad route started at Teraswitch’s Miami facility. From there it propagated to data centers in London, Amsterdam, Frankfurt, Singapore, and Tokyo. North America was unaffected. Teraswitch corrected the route in about 10 minutes. Traffic was flowing again by 4:16 a.m. UTC. Most affected validators, including the large operator Helius, remained down for the full 33 minutes because their backup systems never engaged.
Concentration is what made the incident severe. One network operator, identified as AS2032, controlled more than a quarter of all staked tokens securing Solana. That already exceeded the network’s own prescribed safety limit. Almost all of those tokens went dark at once. Other companies lost another 14 million tokens in the same short window.
The practical upshot: a single autonomous system carrying outsized stake failed, and the network’s safety margin nearly vanished with it.
Roughly 90 validators were affected. They lost 333 SOL in rewards, a sum Marinade called relatively small and said would be covered by “validator bonds.” No network outage resulted. Neither source reports transactions reversed or affected.
Solana ranks among the leading smart contract blockchains. Assets worth $4.3 billion sit in DeFi protocols on the network. This episode is the closest the chain has come to a finality loss since a five-hour outage in February 2024.
“If delinquency had gone past a third, nothing finalizes for anyone holding SOL anywhere, and there’s no bond for that,” Marinade wrote in its explainer post. “The February 2024 halt took about five hours to restart.”
Neither source reports an official statement or remediation plan from the Solana Foundation beyond Teraswitch’s route fix. The identity of AS2032 beyond the autonomous system number is not named in either report. The exact start time of the incident is not stated; only the 4:16 a.m. UTC resolution is given.
The deeper open question is structural. A single hosting provider’s malformed route pushed a major L1 to the edge of finality loss because too much stake sat behind one network operator. Teraswitch patched the route. The concentration that turned a 10-minute misconfiguration into a near-halt is still in place.
