Polkadot ran degraded for about 14 min
Polkadot ran degraded for about 14 min, from 07:58 to 08:12 UTC on 17 Sep 2024. From the public record: postmortems, status pages and news reports, linked below.
We have measured Polkadot ourselves since 12 Nov 2026; anything earlier is from the public record.
What happened
The runtime 1.3.0 upgrade migrated HostConfiguration at the end of the block, so a runtime API read corrupted values and the candidate-backing subsystem stalled; relay finality lagged about 70 blocks and parachains nearly stopped.
Sources
Polkadot right now
Producing blocks
Compare all chains in the reliability league
How this works and where the data comes from
What we measure. Every minute a server job asks each chain for its newest block and reads the timestamp the chain itself put in that block. "Last block" is now minus that timestamp, so it is accurate even if our own checks run late. "Last hour" is the average spacing between blocks over the past hour, also from block timestamps.
When we call a halt. A chain counts as halted when no new block has appeared for much longer than normal. The limit is set per chain: 2 minutes for most fast chains, 5–7 minutes for Scroll, ZKsync and Cardano, 14 minutes for Dogecoin, 30 minutes for Litecoin and 2 hours for Bitcoin, where long gaps happen by chance. Before declaring a halt we ask a second, independent data provider. If one provider is stale and the other is fresh, the chain is fine and we log the provider as lagging. For Solana, Sui and Cardano we have only one free source, so we wait for three stale checks in a row. Start time is the last block before the stop; end time is the first new block, within about a minute.
Finality. For Ethereum we read the last finalized checkpoint from the beacon chain (normally 2 epochs, about 13–19 minutes behind). For other EVM chains and Polkadot we read the newest "finalized" block from the node. On rollups (Base, Arbitrum, OP, ZKsync, Scroll), "finalized" means settled on Ethereum, which takes longer by design: hours on ZKsync. Free Solana endpoints do not report commitment levels reliably, so we do not show Solana finality.
Limits. We have been measuring since 12 Nov 2026; uptime figures cover only that period and say so. If our job misses minutes, a halt that started and ended entirely inside that gap would show up only as a slow hour in the chart. The historical record before our measurements comes from postmortems and news reports, with approximate times marked "≈". Hyperliquid is measured through HyperEVM, which stops whenever the HyperCore L1 stops.
Information only, not financial advice. Data comes from public sources and can be delayed or wrong; check the primary source before acting.