On the morning of August 12, the Solana network came within a step of halting transaction finality. The cause wasn't a hack, but an ordinary routing failure at a single data center run by hosting provider Teraswitch in Miami. Within minutes, almost 29% of staked SOL went offline. Less than five percentage points separated the network from the threshold where it stops confirming transactions.
What exactly happened to the network?
An internal Teraswitch route, normally used to signal that an edge router can reach the internet, propagated out of Miami without its correct metrics attached. A route reflector in Amsterdam picked it up and pushed it into Europe and Asia. Edge routers in London, Amsterdam, Dublin, Frankfurt, Singapore, and Tokyo treated it as local, even though the traffic actually hit a dead end at the data center core. North America never felt the problem.
Teraswitch engineers found the fault within ten minutes. Traffic recovered at 04:16 UTC, and the whole incident lasted roughly 33 minutes. About 90 validators were affected, losing a combined 333 SOL in rewards. That sum will be covered by so-called validator bonds, collateral operators post in advance for exactly this kind of downtime.
The network halts once a third of stake goes dark
Solana stops finalizing blocks for good once more than a third of staked tokens go offline at the same time. That threshold isn't arbitrary. Most Proof-of-Stake networks rely on the same Byzantine fault tolerance margin. The system keeps running as long as the inactive or malicious share of validators stays below a third of the vote. Cross that mark, and the chain loses the ability to confirm even a single block.
The last time that line was fully crossed was February 2024, when recovery took nearly five hours. That outage came from a transaction queue overwhelmed by a memecoin frenzy, a load problem rather than a provider infrastructure failure. Back then the network needed a coordinated manual validator restart, while this time a full halt was avoided entirely. This incident exposed a different kind of risk. It hides not in the protocol's code, but in the network topology surrounding it.
Why could one provider affect a quarter of the network?
The core problem isn't the routing bug itself, but how concentrated staking turned out to be. Autonomous system AS20326 carried more than a quarter of all staked SOL. That's above the 25% cap set by the Solana Foundation's delegation program. When that system went down, 94% of its linked tokens dropped offline at once, not gradually the way a genuinely distributed setup would fail.
An analysis by staking platform Marinade Finance turned up several more weak spots in the network:
- 59 validators holding 80.2 million SOL simply waited for the route to reconverge instead of switching to backup channels.
- Helius, Solana's second-largest validator, stayed offline for the full 33 minutes because its failover never kicked in.
- Another 14.1 million SOL dropped out through other providers (latitude.sh, Limestone, Butterfly Research, and Allnodes), and the reason for that overlap is still unclear.
- Marinade itself admitted that four autonomous systems hold two-thirds of the stake in its own allocation model, with one alone reaching 36.94%.
That last figure matters most. The platform that spreads delegated tokens across validators turned out to carry the same concentration risk the network had just demonstrated.
Solana Foundation insists the network held up
Solana Foundation VP of Technology Jacob Creech argued in a post on X that the network's design did exactly what it was supposed to. According to him, 597 of 699 staked validators kept voting, and affected participants recovered within 40 minutes. Validators in the Foundation's own delegation program weren't touched at all. Those figures differ slightly from Teraswitch's own account, which claimed traffic was fully restored in just 33 minutes. The gap likely comes from some validators needing extra time to resynchronize after the route itself was already fixed.
"Because Solana validators are distributed across independent infrastructure providers, the failure of a single provider did not interrupt the network. This is decentralization and infrastructure diversity working as intended, and a proof point for Solana's resiliency through periods of stress."
- Jacob Creech, VP of Technology at Solana Foundation, in a post on X, August 12, 2026
SOL's price on exchanges barely moved in response, and observers didn't record any notable capital outflow from the network. The gap in tone between the Foundation's official stance and Marinade's warnings speaks for itself. One side points to proven resilience, while the other flags a risk that simply hasn't fully played out yet.
Should SOL holders actually worry?
Blocks kept getting produced and transactions kept landing throughout the incident. In that sense, Solana, often pitched as a faster, cheaper alternative to Ethereum, did hold up under pressure. But getting 86% of the way to a full halt isn't the kind of safety margin worth counting on regularly, especially with DeFi protocols on Solana currently holding more than $4.3 billion in locked assets.
For traders and holders of DeFi positions on Solana, the takeaway is simple. Stake concentration at a single provider remains a real risk, even though this particular event resolved without damage. Marinade has already promised to revisit its own allocation model. Whether the Solana Foundation follows suit remains an open question. Incidents like this are becoming a more frequent reminder that decentralization at the validator level doesn't always mean decentralization at the level of the physical infrastructure connecting those validators to the network.
Comments
Your email address will not be published. Required fields are marked *