Starknet crashed again on Monday, this time it took 18 minutes to recover.

The cause was an "memory confusion" in the execution layer — the system incorrectly retained data that should have been deleted, leading to deviations in transaction execution. Fortunately, the proof layer detected the anomaly in time and prevented the problematic data from being committed on-chain, but the cost was a forced reorganization of the entire block, causing the network to roll back.

Honestly, this has become a routine operation this year. Remember the incident in September? A Sequencer failure triggered a chain reaction, causing over five hours of downtime, two reorganizations, and transaction records to be wiped out. For ordinary users, it might just be a minor annoyance, but for those doing high-frequency trading and emergency stop-loss, it’s a real asset risk.

Layer2’s original goal was to be fast and cheap, but now it seems that even basic stability is still compromised. Every time there’s an incident, the official promises to strengthen testing and improve auditing processes, but frequent rollbacks and interruptions make it hard to see hope.

This situation has a pretty deep impact on the entire Layer2 ecosystem. Users will start to consider: what’s the use of fast speed if stability is problematic? How can security be guaranteed? Have you experienced transaction delays or rollbacks on Starknet? Are these repeated interruptions changing your expectations of Layer2?

Would love to hear everyone’s real experiences.