Summary

Completes journal-based sync by adding a journals strategy to the node: instead of executing host blocks, a syncing node subscribes to upstream journal WebSocket sources, applies the pre-computed state diffs to storage, and hands off to live block execution once it catches up to the host tip. Builds on the in-process journal chain and JournalHashes persistence from the downstack commits.

Changes

• Add SyncStrategy (blocks default, journals) and journal-sync settings to JournalConfig: upstream source URLs and client stall-timeout/backoff, with SIGNET_JOURNAL_* env vars, cross-field validate(), and startup misconfiguration warnings.
• New journal_sync module. JournalIngestor applies journal-chain events to hot/cold storage - journal application and reorg unwind - and broadcasts newHeads/reorg notifications to RPC subscribers. runtime owns client/ingestor setup and lifecycle, the run-loop, journal-client construction, and checkpoint seeding from the JournalHashes table so a restarted node resumes from its storage tip rather than re-bootstrapping from genesis.
• Wire the strategy into the node. run_journal_sync spawns the journal client and ingestor alongside the chain, applies journals until the rollup tip is within a margin of the host tip, then hands off to run_block_sync for live execution. Block sync is extracted into its own method and gains a cancellation path; a new cancellation token drives graceful shutdown.
• Extend HostNotifier with host_tip() (used to decide when the node has caught up) and backpressures_host() (default false; the reth ExEx returns true); implement both for the reth and RPC backends.
• Drain and discard host notifications during journal sync for backends that backpressure the host (the reth ExEx). Because the node runs in-process as a reth ExEx, leaving notifications unconsumed would fill reth's buffer and stall its pipeline; draining keeps the host moving while state is derived from the journal feed. Finished-height is never signalled, so the host retains the blocks the post-catch-up handoff backfills.
• Require with_cancellation_token on SignetNodeBuilder.
• End-to-end tests: journal application, transition at startup and after catch-up, source exhaustion, reorg replay, restart-resume from persisted checkpoints, and host-notification draining. The test harness gains a shared host-tip counter and a backpressure-emulating notifier.

Notes

• The journals strategy assumes the host (reth) is already at tip when the ExEx is enabled. During reth's own pipeline backfill an un-draining ExEx stalls the host; drain-and-discard mitigates this, and operationally the ExEx is enabled only after reth syncs (SIGNET_DISABLE_EXEX).
• Journal-synced blocks carry state but no transaction/receipt/event history, so RPC queries over the journal-synced range return empty tx data.
• Downstream signet-exex must add .with_cancellation_token(...) when bumping to this version.
• ENG-2016.

🤖 Generated with Claude Code