Solana came closer to a serious network disruption on August 12 than initial reports suggested. A routing failure at infrastructure provider Teraswitch knocked 28.83% of all staked SOL offline for roughly 33 minutes, pushing the network within 4.51 percentage points, or about 19.9 million SOL, of the 33.34% threshold at which Solana would stop finalizing transactions altogether. The Solana network continued producing blocks throughout the incident, and validators in the Delegation Program weren't affected, but the margin to an actual halt was thinner than the “102 of 699 validators” framing that first circulated made it sound.
Marinade Finance's analysis of the incident pointed to a deeper structural issue behind the numbers: one single autonomous system was found to be hosting about 118.9 million SOL, more than a quarter of all staked SOL on the network, and roughly 94% of that concentrated stake went offline together when Teraswitch's routing broke. That concentration, more than the raw validator count, is what pushed the network as close to a halt as it got.
What Actually Broke
According to Teraswitch's own incident status page, the fault traced back to a malformed default route advertised from its Miami (MIA1) facility, which propagated across the provider's backbone through a route reflector in Amsterdam and caused simultaneous disconnections at twelve sites across Europe and Asia-Pacific. Engineers identified the malformed route within 10 minutes and removed the Miami site from the backbone to stop it spreading further; affected sites reconverged on valid local routes and service was fully restored at 04:16:15 UTC.
The Cost of 33 Minutes
The roughly 90 affected validators collectively lost 333 SOL in rewards during the outage window, a relatively modest direct cost set against how close the broader network came to a more serious disruption. Teraswitch has since pushed a global configuration change across its compute sites intended to prevent a similarly malformed route from blocking traffic forwarding again, though the provider says a full root-cause report is still pending.
