On August 12, the Solana network found itself on the brink of a critical failure. Due to routing issues at the infrastructure operator TeraSwitch, validators controlling 28.83% of all staked SOL went offline simultaneously. This is just 4.5 percentage points below the 33.34% threshold at which the blockchain loses its ability to finalize transactions and effectively comes to a halt.
The incident affected approximately 90 validators. TeraSwitch's autonomous system (AS20326) accounts for 118.9 million SOL, or 27.34% of the network's total stake. At the moment of the failure, 94% of that amount went offline—a rare case where infrastructure concentration creates systemic risk for the entire blockchain.
Fortunately, the critical level was not reached. Routing was restored in about 33 minutes, and the validators came back online. However, during that time, they missed out on approximately 333 SOL in rewards—a small amount, but the signal itself is alarming.
The failure exposed a chronic decentralization problem
At Marinade Finance, one of the largest liquid staking protocols in the ecosystem, the event was described as a striking example of infrastructure concentration risk. The project has already announced a review of limits on the distribution of validators across autonomous systems, data centers, and backup capacities.
It's important to understand: this problem is not new. As early as July 22, TeraSwitch was servicing validators with 27.1% of locked tokens. It was followed by UAB Cherry Servers (12.7%) and Latitude.sh (11%). Such concentration makes the network vulnerable to single points of failure, which contradicts the very philosophy of decentralized blockchains.
Some major players have already recognized the risks. For example, Coinbase reported in its first-quarter report that 13 of its validators operate through TeraSwitch, and another 10 through Latitude. At the same time, for each operator, the exchange maintains a backup server in a different location to minimize the consequences of a single provider's failure.
History knows even more serious precedents. In November 2022, German hosting provider Hetzner shut down servers with Solana nodes, affecting about 40% of validators with roughly 20% of staked coins. The network survived then, but the question of resilience remained open.
I'll add that the number of active Solana validators dropped to 800 in January 2026—the lowest since 2021—whereas at the peak in 2023, there were more than 2,500. This only exacerbates the problem: the fewer validators there are, the higher the concentration of risks on individual operators.
My conclusion: the TeraSwitch incident is not just a technical failure, but a wake-up call for the entire Solana ecosystem. If major stakers and protocols do not begin to diversify infrastructure more actively, the next similar case could end not with 33 minutes of downtime, but with a complete network halt. The market has not yet reacted appropriately, but the consequences for trust in the network could be long-term.