OP Mainnet external nodes and RPC providers stalled on a stale head (7 July 2026)
Status: Resolved
Impact
External nodes and third-party RPC providers were stuck at a stale head for ~22 minutes (17:59 to 18:22 UTC). Because major providers were serving that stale head, transaction throughput was near zero for about 16 minutes of that window, and submissions routed through those providers did not land. Reads continued to be served, though against a stale state. There was no reorg, no safe-head or finality violation, and no funds were at risk.
Root cause
A routine infrastructure upgrade disrupted a network connection our sequencers depend on. The connection failed silently rather than erroring, which throttled block production and pushed new blocks outside the window for normal peer-to-peer propagation. External nodes recovered once the affected range was posted to L1
Remediation
We deployed a fix removing the dependency responsible for the slowdown and made the affected connections fail cleanly when infrastructure cycles. We're also strengthening internal alerting and node peering so issues of this kind are detected and contained faster.
