Starknet’s most ambitious upgrade to date has gone live, but it promptly reminded users of the risks of pushing cutting-edge infrastructure into production.
The layer-2 network rolled out its Grinta upgrade earlier this month, a sweeping release that touches nearly every layer of its architecture. Among the headline features:
- Instant transactions with ~0.5 second pre-confirmations, bringing a “Web2-like” user experience.
- Seven-times faster block production, reducing latency for DeFi, gaming, and AI applications.
- A new fee market modeled on Ethereum’s EIP-1559, with STRK as the default gas token. Paymasters can still let users pay in ETH or stablecoins.
- The first step toward sequencer decentralization, with three sequencers sharing responsibility instead of a single centralized one.
- A mempool upgrade that lets users prioritize transactions by fee, akin to Ethereum’s gas bidding.
- A standardized Paymaster API (SNIP-29) to simplify integration for wallets and dApps.
StarkWare, the team behind Starknet, framed the release as a historic milestone. “Starknet is moving from a centralized network to a credibly neutral, performant network ready to onboard millions of users,” the project said in its Grinta announcement.
Outage overshadows launch
But the triumph was quickly tempered. On September 2, just after Grinta’s debut, Starknet went offline for more than two hours. A chain reorganization wiped transactions submitted during a roughly two-hour window, frustrating users and raising questions about rollout strategy.
According to the incident report published today, the root cause was “ungraceful recovery from earlier P2P issues on the Ethereum network.” Specifically, the sequencer processed an unprovable stream of transactions, forcing a rollback.
Developers acknowledged the severity: “It sucks, plain and simple,” StarkWare contributor Abdelhamid Bakhta wrote on X. He emphasized, however, that “no one has decentralized a ZK rollup architecture on a live chain before. Setbacks come with the territory, but they drive progress.”
The upgrade’s scale contributed to the risk. StarkWare batched multiple features into a single release, rather than phasing them in. “Usually when we know there is a major upgrade coming that will anyway require a short downtime, we prefer to batch multiple features,” Bakhta explained in response to community criticism. Strategically, he added, the team wanted Grinta live ahead of “something important coming very soon related to the Bitcoin plans.”
Reaction was mixed. Bakhta compared the episode to Solana’s repeated outages — a network that nevertheless retained developer momentum. “Bullish on Starknet downtime. Solana playbook, all planned,” he quipped. Others noted the irony of a rollup, whose very selling point is Ethereum-grade reliability, suffering downtime.
Researcher Eric Wall, a StarkWare advisor, pressed why the rollout wasn’t staged more carefully. Bakhta countered that progress in uncharted ZK rollup design comes with inevitable turbulence.
Despite the hiccup, Grinta marks a turning point. Starknet is now the first rollup to decentralize sequencing in production, technically, although all sequencers operated by StarkWare for now. Future phases will bring in independent operators, aiming to improve censorship resistance, reliability, and throughput.
As the outage fades from memory, the scope of the upgrade should come to the fore. As Bakhta put it: “We’re humbled by the issues, but proud of what we’ve achieved. Let’s fix it, learn from it, and keep building.”
Get the news in your inbox. Explore Blockworks newsletters: