Starknet, the $420 million ZK-rollup, released a post-mortem detailing the root cause of its January 5 network halt. The incident forced the protocol to trigger a block reorganization, effectively erasing 18 minutes of confirmed activity to prevent invalid state transitions from reaching Ethereum mainnet.
The Disconnect: Execution vs. Proving
The outage stemmed from a logic error in the blockifier, Starknet’s execution engine. According to the engineering report, the blockifier incorrectly retained state-writes from a reverted function during a specific sequence of cross-contract calls. While the sequencer accepted the invalid transaction, Starknet’s proving layer, the cryptographic mechanism that verifies validity, correctly rejected it.
This discrepancy created a deadlock. The sequencer considered the block valid; the prover refused to generate a proof for it. To resolve the mismatch, the team paused the sequencer and reorganized the chain, dropping all transactions processed between 09:24 and 09:42 UTC.
In one specific combination of cross-function calls… the blockifier remembered a state-writing that happened within a function that was reverted.
The network remained offline for approximately five hours while engineers deployed a hotfix.
Institutional Context
This marks Starknet’s second major liveness failure in four months. In September 2025, the network suffered a nine-hour outage following the ‘Grinta’ upgrade (v0.14.0), which also required chain reorganizations due to RPC and sequencer failures. Recurring mismatches between execution clients and proving engines highlight the fragility still present in high-throughput ZK-rollups.
Markets shrugged off the technical debt. STRK traded flat at $0.082 (-1.6%), with volume remaining consistent despite the operational pause.
