fix: replace panic! with graceful error return in commit_shard_chunk and commit_block

[Sertug17]

Summary

Fixes issue #1015.

Both commit_shard_chunk (src/storage/store/engine.rs) and commit_block (src/storage/store/block_engine.rs) called panic!() when replay_proposal returned an error. This crashes the entire node process a single transient storage error would take down a validator or full node completely.

Problem

These functions are called after consensus has already decided on a value. At that point the node is committed to applying the block. If replay_proposal returns an error (e.g. transient I/O failure, unexpected DB state), the previous code panicked:

// BEFORE (engine.rs:2722 and block_engine.rs:2248)
Err(err) => {
    error!("State change commit failed: {}", err);
    panic!("State change commit failed: {}", err); // crashes entire node
}

This means:

• Any transient storage error causes the node to go completely offline
• The node falls behind the network and must resync from scratch
• In production this can cause extended outages that are hard to diagnose

Fix

Replace panic!() with early return, keeping the existing error!() log:

// AFTER
Err(err) => {
    error!("State change commit failed: {}", err);
    return; // node stays alive, error already logged
}

The error is already logged clearly. The node stays alive and can continue processing future blocks rather than crashing.

Files Changed

• src/storage/store/engine.rs commit_shard_chunk
• src/storage/store/block_engine.rs commit_block

Closes #1015

[vercel]

@claude is attempting to deploy a commit to the Internal - Neynar Team on Vercel.

A member of the Team first needs to authorize it.