> For the complete documentation index, see [llms.txt](https://docs.propw.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.propw.com/blog/solana-nearly-hit-a-network-halt-the-infrastructure-risk-behind-the-networks-near-miss.md).
# Solana Nearly Hit A Network Halt: The Infrastructure Risk Behind the Network’s Near-Miss
One routing error at a hosting provider called Teraswitch knocked 28.83% of Solana's staked SOL offline on August 12. The network came within 4.5 percentage points of the 33.34% threshold where it stops finalizing transactions. About 20 million SOL stood between a functioning chain and a freeze.[ Marinade Finance](https://www.tradingview.com/news/beincrypto:f90d61f6a094b:0-solana-network-nearly-stopped-working-today-should-sol-investors-worry) reconstructed the incident and called it getting "86% of the way to a halt."
The market barely noticed. SOL ticked up 0.6% on the day. No user funds were at risk. The 90 affected validators lost 333 SOL in rewards, roughly $25,600, covered by validator bonds. But the structural picture is what matters. Solana processes 171.9 million non-vote transactions per day and just crossed its first billion-transaction week. Its consensus backbone nearly froze because one data center in Miami misconfigured a route.
### Layer One: Infrastructure Concentration
The failure started at Teraswitch's Miami site. A bad route spread through an internal relay in Amsterdam. Twelve locations from London to Tokyo lost connectivity. North America felt nothing. The fix took 10 minutes to identify and 33 minutes total to resolve.
The damage concentrated in one autonomous system number. AS20326 hosts 118.89 million SOL, or 27.34% of all staked supply on Solana. During the fault, 94% of that stake went offline simultaneously. The Solana Foundation Delegation Program caps any single ASN at 25% of delegated stake. That cap exists for exactly this scenario. It is already broken.
Only 3 of 74 validators Marinade measured successfully switched to backup infrastructure: Laine, Cogent Crypto, and Lion3d. The remaining 71 sat offline until routing reconverged. Helius, Solana's second-largest validator by stake weight, stayed down for the full 33 minutes. Marinade acknowledged its own delegation program has the same problem, with four ASNs holding two-thirds of its allocated stake.
Geographic distribution does not equal infrastructure independence. Validators in London, Amsterdam, Dublin, Frankfurt, Singapore, and Tokyo all went dark because they shared the same upstream provider. A blockchain can list 715 validators across six continents and still concentrate two-thirds of its voting power behind one router in Miami.
### Layer Two: Validator Stake Concentration
A July 22 on-chain snapshot recorded[ 715 validators and 426.4 million SOL of active stake](https://www.analyticsinsight.net/news/solana-validators-challenge-new-sol-tokenomics-proposals-what-investors-need-to-know). The top 10 validators controlled 24.08% of that stake. The top 100 controlled 72.86%. Just 18 validators collectively cross Solana's one-third superminority threshold, the minimum number of entities that could theoretically collude to halt consensus.
Anatoly Yakovenko, Solana's co-founder,[ acknowledged](https://almerirja.vercel.app/?live-blog-18110076-2026-07-21-back-to-the-list-solana-co-founder-yakovenko-unveils-nakamoto-vision-after-ai-ro/) in July that the network's Nakamoto coefficient sits near 20. For comparison, Ethereum's estimated Nakamoto coefficient exceeds 7,000. Yakovenko framed the gap as a long-term project, drawing a parallel to the time between the American Revolution and the Constitution. The Teraswitch incident suggests the timeline for closing that gap may be shorter than the rhetoric implies.
The staking layer compounds the problem. Jito's Block Assembly Marketplace has grown to[ 378 validators and 33% of Solana's total stake weight](https://solanacompass.com/news/jito-q2-2026-protocol-revenue-falls-45-to-128m-as-bam-reaches-33-of-solana-stake), up from 27.7% in Q1. The broader Jito client family covers roughly 54% of active stake. Concentration in staking infrastructure means concentration in MEV extraction, block production, and consensus influence.
### Layer Three: Client Software Concentration
Client diversity is the third axis, and Solana is making progress but remains top-heavy.[ Jito-Solana, a variant of the Agave client](https://www.soladex.io/glossary/client-diversity), runs on roughly 70% of staked validators. The Firedancer family, including Frankendancer, holds about 20%. The vanilla Agave client covers the remainder in single digits.
Jito-Solana is a thin fork of Agave designed primarily for MEV-focused block production. It shares much of Agave's codebase, meaning bugs in one frequently appear in the other. Firedancer, written from scratch in C by Jump Crypto, reached v1.0.0 in June 2026 and shares zero code with the Agave lineage. The ecosystem's stated goal is to push 50% of stake onto Firedancer by Q4 2026. That would meaningfully change the risk picture. Until then, a single bug in the Agave codebase can affect 90% of the network.
Ethereum learned this lesson the hard way. It runs four independent execution clients with a standing rule that no single client should exceed 66% of stake. Solana is running the same playbook approximately two years behind, with outages as its primary failure mode rather than invalid state finalization.
### The Tokenomics Squeeze
The near-miss arrives as validators face a governance vote that could thin their margins.[ SIMD-0550](https://solanacompass.com/news/solana-validators-debate-simd-0550-and-simd-0553-as-sol-issuance-cuts-near) would double Solana's annual disinflation rate from 15% to 30%, pulling the 1.5% terminal inflation floor forward from 2032 to 2029. Staking yield drops from 5.84% to 4.34% in the first year, then to 3% by year two and 2.25% by year three. SIMD-0553 would replace the flat per-signature fee with a resource-based charge that is 100% burned, lifting daily SOL burns from roughly 648 to between 7,500 and 9,000.
The discussion window closes August 22. Passage requires two-thirds of participating stake. DeFi Development Corp., which holds SOL as its primary reserve asset, has pledged support. Smaller validators are pushing back. During the SIMD-0550 review, one participant noted that validators near the break-even threshold face "immediate extinction" if rewards fall faster than fee revenue grows.
The paradox is straightforward. Tokenomics reforms that reduce issuance are supply-positive for SOL holders. They are also centralizing forces for the validator set. Smaller operators running on thin margins exit first. The validators that remain hold more stake. The Nakamoto coefficient moves in the wrong direction.
### What Happens If the Next One Crosses 33.34%
Solana's last full network halt was February 6, 2024. It took five hours to restart. No validator bond covers that outcome. A halt freezes every SOL holder, every DeFi position, every pending transaction simultaneously.
The Alpenglow upgrade, targeting 100 to 150 millisecond finality, is due by October. It prioritizes safety over liveness during faults, meaning it would halt rather than produce uncertain blocks. That is the right design choice for correctness. It also means the next routing failure that crosses the threshold produces a longer freeze, not a shorter one.
If Solana's stake distribution remains unchanged and another ASN-level failure removes 28 to 30% of stake, the network enters a state where blocks may continue producing but transactions cannot finalize. Applications depending on irreversible settlement, which is most of DeFi, stop functioning. The market reaction would not be a 0.6% uptick.
The questions worth watching are whether Marinade follows through on tighter ASN caps and public failover reporting, whether the SIMD-0550 vote accelerates small-validator attrition, and whether Firedancer adoption reaches 50% before the next bad route finds a concentrated stake cluster. Solana's throughput is not in doubt. Its consensus distribution is.
### About PropW
PropW is a world-leading Web3-focused proprietary trading platform specializing in digital assets. Designed to bridge the gap between ambitious traders and institutional-grade liquidity, PropW provides structured trading challenges and advanced evaluation systems. Our mission is to identify, fund, and scale top-tier trading talent globally, offering users the opportunity to manage up to $200,000 in funded capital while retaining the majority of the profits.
For more information, visit: [Website](https://www.propw.com) | [Twitter](https://x.com/PropWGlobal) | [Telegram](https://t.me/PropWGlobal) | [LinkedIn](https://www.linkedin.com/company/propw) | [Discord](https://discord.com/invite/C8enuujTrZ)
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://docs.propw.com/blog/solana-nearly-hit-a-network-halt-the-infrastructure-risk-behind-the-networks-near-miss.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
