feat: add startup.defer-gossip flag to ensure nodes restart status

[sysvm]

Description

When upgrading or restarting RPC nodes, the nodes need about 30 minutes to sync blocks which may cause node unstability.

Rationale

Defers P2P gossip processing during startup until op-geth's unsafe head has caught up to the live tip via L1 derivation. This avoids the driver/alt-sync activity loop that occurs when an RPC node restarts with a large unsafe-head gap. Recommended for RPC and bridge nodes; not needed for sequencer and P2P nodes.

Example

Add --startup.defer-gossip flag in RPC nodes.

Changes

Notable changes:

• N/A

[hashdit-bot]

Pull Request Review

This PR introduces a new optional flag, --startup.defer-gossip (env: OP_NODE_STARTUP_DEFER_GOSSIP), and wires it through configuration to control startup behavior in op-node. When enabled, the node now waits for op-geth’s unsafe head to catch up (with polling, lag threshold, and max timeout) before fully accepting gossip, and temporarily drops most incoming unsafe payload gossip during that catch-up window. It also adds a one-time payload allowance during catch-up to unblock EL sync state transitions and then restores normal gossip flow once the gate is opened.

Sensitive Content

No sensitive content detected.

Security Issues

No serious security issues detected.

---

Generated by Hashdit Bot. This tool can absolutely NOT replace manual audits.