Solana outages are the network’s most cited weakness, and the honest record is that mainnet beta has fully halted roughly seven times since launching in 2020 — most of them clustered in 2021 and 2022. Two were caused by floods of bot transactions overwhelming validators; the rest were software bugs in the validator client that required operators to coordinate a manual restart. The last full halt was on 6 February 2024, which means that by August 2026 the chain has run continuously for about two and a half years, its longest stretch by a wide margin.
Key Facts
- Around seven full halts since mainnet beta launched in March 2020, plus several periods of severe degradation that were not complete stoppages.
- Two outages were triggered by transaction spam; the remaining five came from bugs in the validator client.
- Longest single outage: roughly 19 hours starting 25 February 2023, caused by an oversized block breaking block propagation.
- Most famous: 14 September 2021, about 17 hours, after bot traffic around a Raydium IDO exhausted validator memory.
- Last confirmed halt: 6 February 2024, roughly five hours, from a bug in the legacy program loader.
- No full network halt since then — approximately two and a half years of continuous block production as of August 2026.
- A halt requires validators to agree on a restart slot offline, which is why recovery is measured in hours, not minutes.
The Full Record
| Date | Approx. duration | Cause | Category |
|---|---|---|---|
| December 2020 | ~6 hours | A block that some validators could not validate split consensus | Client bug |
| 14 September 2021 | ~17 hours | Bot traffic around the Grape IDO on Raydium exhausted validator memory | Spam |
| 30 April – 1 May 2022 | ~7 hours | NFT minting bots flooded the network, consensus lost | Spam |
| 1 June 2022 | ~4.5 hours | Durable nonce implementation bug | Client bug |
| Late September 2022 | ~6 hours | A hot-spare validator produced duplicate blocks; fork choice logic could not resolve them | Client bug |
| 25 February 2023 | ~19 hours | A malfunctioning node broadcast an oversized block, overwhelming the Turbine propagation layer | Client bug |
| 6 February 2024 | ~5 hours | Bug in the legacy program loader caused cascading failures | Client bug |
January 2022 also brought a week of severe congestion that gets counted as an outage in some retellings. It was not one — blocks kept being produced, but a large share of transactions failed to land, which for a user is barely a distinction. That episode did more than any halt to push through the priority fee mechanism that now governs how block space gets allocated.
Two Different Failure Stories
The spam outages of 2021 and 2022 shared a shape. Solana had no meaningful pricing on block space and no way to tell a genuine transaction from the eight hundredth retry of a bot, so a hot launch turned into an undifferentiated flood that pushed nodes past their memory limits. The fixes were structural: QUIC transport with per-sender connection limits, stake-weighted quality of service so traffic from staked validators takes priority, and local fee markets that let one saturated pool get expensive without dragging the whole chain with it.
The bug outages were different and, in a sense, more serious. Each traced to a single implementation flaw in a codebase every validator on the network was running, meaning one defect could stop everything simultaneously. That is a monoculture problem, not a capacity problem, and no amount of fee market design addresses it. The answer is client diversity — which is the whole strategic point of a second independent validator client, since a bug in one implementation should degrade the network rather than stop it.
Why Restarts Take Hours
When Solana halts, it does not resume on its own. Validator operators — hundreds of independent teams across dozens of countries — have to agree on which slot to restart from, download a matching snapshot, and bring nodes back with enough combined stake to reach supermajority. That coordination happens over Discord and calls, at whatever hour it happens to be, which is why a five-hour outage often reflects thirty minutes of technical work and four and a half hours of humans reaching consensus about the technical work. It is also, uncomfortably, evidence of decentralisation: a chain that could be restarted by one party in ten minutes would tell you something worse.
Where Things Stand in 2026
Two and a half years without a halt is a real result, not a marketing line, and it covers a period including memecoin manias that would have flattened the 2021 network. The caveat worth keeping is that the underlying risk profile has changed rather than vanished: the spam class of failure is genuinely engineered out, while the client bug class depends on how far Firedancer adoption gets. Judge it on evidence rather than either narrative — uptime trackers, block production history on Solscan, and value actually parked on the chain via DefiLlama say more than any think piece about whether Solana is finished.
What Outage Risk Means for Holding $DOLAN
DOLAN Duck ($DOLAN) has a fixed 98.3M supply spread across roughly 10,700 holders, and a network halt would not touch a single unit of it. Balances are ledger entries; when the chain stops, they stop with it, frozen exactly as they were. The real exposure is liquidity, not custody — during a halt nobody can sell, and when blocks resume everyone tries to at once, so the first minutes after a restart are historically the most violent price action a token sees. That risk lands hardest on leveraged and time-sensitive positions, and barely at all on someone holding spot in their own wallet. It is one more argument for treating who actually runs the network as due diligence rather than trivia.
Roughly seven full halts since mainnet beta launched in 2020, concentrated in 2021 and 2022. Two were caused by transaction spam and five by bugs in the validator client software.
6 February 2024, when a bug in the legacy program loader halted the chain for about five hours. As of August 2026 that remains the most recent full outage.
The 25 February 2023 halt, at roughly 19 hours. A malfunctioning node broadcast an unusually large block that overwhelmed Turbine, Solana’s block propagation protocol.
Nothing is lost. Balances are ledger entries and simply freeze in place until block production resumes. The practical risk is being unable to trade, not losing tokens.
Because restarting requires hundreds of independent validator operators to agree on a restart slot, load matching snapshots and come back online with enough combined stake. Most of the delay is coordination, not repair.
Spam-driven outages are largely addressed through QUIC transport, stake-weighted quality of service and local fee markets. Client bug outages depend on running more than one independent validator implementation, which is still in progress.
Not since February 2024, giving roughly two and a half years of continuous block production as of August 2026 — the network’s longest stability streak so far.