Relay submissions more metrics

[dvush]

📝 Summary

• add more metric for relay submissions
• move some blocking code out of tokio
• relay submit: use gzip fast compression instead of the default
• add e2e submission metrics to test relay

✅ I have completed the following steps:

• Run make lint
• Run make test
• Added tests (if applicable)

[Copilot]

Pull Request Overview

This PR enhances metrics collection for relay submissions by adding timing and compression metrics, moving blocking operations out of Tokio's runtime, and optimizing gzip compression for better performance.

• Add more detailed metrics for relay submission timing and payload sizes
• Move blocking compression operations to dedicated blocking threads
• Switch from default to fast gzip compression for relay submissions

Reviewed Changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
crates/test-relay/src/relay.rs	Add timing metrics for measuring latency from builder to relay
crates/test-relay/src/metrics.rs	Add new histogram metric for time-to-receive measurements
crates/test-relay/src/main.rs	Add periodic histogram reset functionality
crates/rbuilder/src/utils/mod.rs	Add utility function for microsecond timestamps
crates/rbuilder/src/telemetry/metrics/mod.rs	Add new metrics for relay submission steps and payload sizes
crates/rbuilder/src/primitives/mev_boost.rs	Update relay client interface to return submission statistics
crates/rbuilder/src/mev_boost/submission.rs	Add sealed timestamp to bid metadata
crates/rbuilder/src/mev_boost/mod.rs	Implement detailed timing metrics and move compression to blocking thread
crates/rbuilder/src/live_builder/order_input/mod.rs	Move blocking orderpool operations to spawn_blocking
crates/rbuilder/src/live_builder/block_output/relay_submit.rs	Add submission stats metrics collection
crates/rbuilder/benches/benchmarks/mev_boost.rs	Add missing import for benchmark

[Copilot]

The spawn_blocking call moves the entire body_data Vec to the blocking thread, which could be inefficient for large payloads. Consider using a streaming approach or at least moving only the encoder setup to the blocking thread while keeping the data processing async.

[Copilot]

[nitpick] The double question mark operator (??) is used to handle both the spawn_blocking error and the compression error. This creates nested error handling that could be clearer if split into separate statements or using a more explicit error handling approach.

Suggested change

	.map_err(|e| SubmitBlockErr::RPCSerializationError(e.to_string()))??;
	.map_err(|e| SubmitBlockErr::RPCSerializationError(e.to_string()))?;
	body_data = join_result?;

[Copilot]

The spawned task for histogram reset runs indefinitely without proper cancellation handling. Consider adding cancellation token support or at least a way to gracefully shut down this background task.

Suggested change

	tokio::spawn(async {
	tokio::time::sleep(HISTOGRAM_RESET_TIME).await;
	reset_histogram_metrics_test_relay();
	let cancellation = global_cancellation.clone();
	tokio::spawn(async move {
	tokio::select! {
	_ = tokio::time::sleep(HISTOGRAM_RESET_TIME) => {
	reset_histogram_metrics_test_relay();
	}
	_ = cancellation.cancelled() => {
	// Graceful shutdown: do nothing, just exit the task
	}
	}

[Copilot]

The early return inside the spawn_blocking closure prevents the rest of the cleanup logic from running when state provider fails. This could lead to incomplete orderpool maintenance. Consider using a Result return type instead of early returns.