SOL routing bug knocks 29% stake offline

TLDR

A routing bug at a Solana infrastructure provider briefly knocked nearly 29% of staked SOL offline, pushing the network close to a finality halt but it remained operational.

1. Solana (SOL) saw about 28.83% of staked tokens go delinquent after a Teraswitch routing failure, coming within roughly 4.5 percentage points of the halt threshold.
2. The incident exposed heavy stake concentration in a single autonomous system and limited failover among validators, highlighting non obvious centralization and infrastructure risks.
3. Next, watch whether validators diversify providers, improve failover, and how Solana governance and client diversity respond to reduce similar correlated outages.

Deep Dive

1. What Actually Happened

Staking protocol Marinade reported that 28.83% of all staked SOL went delinquent for about 33 minutes when a routing fault at data center provider Teraswitch cut connectivity to multiple validator sites in Europe and Asia. This left Solana within roughly 4.51 percentage points of the 33.34% delinquency level at which the chain stops finalizing transactions, meaning blocks would no longer become irreversible for users if more stake dropped offline, according to Marinade and follow up coverage from Decrypt and The Defiant.

The bug stemmed from a misconfigured default internet route at Teraswitchs Miami facility that was propagated via a route reflector in Amsterdam, causing edge routers to prefer an invalid path and stranding 12 locations without a usable route. Engineers isolated Miami and restored service within minutes, and Solanas status page did not record a full mainnet incident, since block production and transaction processing continued during the episode.

2. Hidden Centralization And Risk

Marinades analysis showed the outage was highly concentrated. One autonomous system, operated by Teraswitch, carried about 118.9 million SOL, more than one quarter of all staked supply and above the Solana Foundations informal 25 percent safety cap, with roughly 94 percent of that stake going dark together. In total, around 90 validators were affected and lost an estimated 333 SOL in rewards, which will be covered by validator bonds.

This clustering means that a single connectivity providers routing bug pushed the network most of the way to a finality halt. Additional stake at other hosting providers also went offline in the same window, suggesting that counting stake merely by cloud or data center brand can understate correlated infrastructure risk.

3. What To Watch Next

Several follow ups are already in motion. Teraswitch has deployed configuration changes and promised a full root cause report, while Marinade signaled plans to tighten stake concentration limits by autonomous system and to disclose validator failover capabilities. Solana ecosystem teams are also progressing on resilience initiatives such as alternative validator clients like Firedancer and performance upgrades to Agave, which reduce dependence on a single software stack.

For users and builders, useful signals will be whether stake distribution across providers becomes less top heavy, whether validators invest in real failover rather than waiting for routing to reconverge, and whether Solana governance explicitly enforces or strengthens infrastructure diversity rules.

Conclusion

Solana narrowly avoided a serious finality event, showing that its consensus design can survive a large infrastructure failure but also revealing that connectivity and stake concentration remain critical weak points. If the ecosystem follows through on diversifying validator infrastructure and enforcing concentration limits, similar routing bugs should be less likely to bring the network so close to a freeze, improving the risk profile for long term SOL holders and applications.