apollo_config,apollo_config_manager: stop re-logging the config load on every poll

[matanl-starkware]

What

ConfigManagerRunner reloads the config on a timer. Each tick re-emitted three info! lines in every pod:

• config_manager_runner.rs — "ConfigManagerRunner: periodic check triggered, updating config"
• loading.rs — "Ignoring default values by overriding with an empty map."
• loading.rs — "Loading custom config file: ..." (once per file)

Now:

• the two loading.rs lines report at info! for the first load of the process and trace! afterwards, via a CONFIG_LOADED_BEFORE atomic;
• the tick heartbeat drops to trace!.

Why

Part of a round of log-cost reduction. Measured over 24h of Mainnet logs via Log Analytics, these three lines are ~9.8% of sequencer-gateway's log bytes and ~3% of sequencer-mempool's — roughly $60/month fleet-wide. They fire in every pod, forever, for a config that almost never changes.

update_config already compares old vs new and no-ops when nothing changed, but load_and_validate_config underneath it logged unconditionally.

Why not just debug!

Production runs RUST_LOG=debug,cairo_vm=warn, so demoting to debug! would save nothing. trace! is below the deployed threshold.

What's preserved

• The startup config load still logs in full at info! — that is the occurrence anyone actually reads.
• Real config changes are still reported by log_config_diff ("ConfigManagerRunner: {key} changed from {old} to {new}") and "Successfully updated dynamic config", both untouched.
• The filesystem-watcher path ("file change detected") is untouched — it only fires on an actual change.
• The flag is only set once a load runs to completion, so a first load that fails early still logs at info! on the next attempt.

Test plan

• cargo build -p apollo_config -p apollo_config_manager
• SEED=0 cargo test -p apollo_config — 34 passed, 0 failed
• SEED=0 cargo test -p apollo_config_manager — 15 passed, 0 failed
• scripts/rust_fmt.sh

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Logging-only change with no behavior to config loading, validation, or update logic; real config changes remain at info.

Overview
Cuts repeated info! noise from periodic config reloads while keeping the first load visible at info!.

In apollo_config, a process-wide CONFIG_LOADED_BEFORE flag and log_config_load route the “ignoring default values” and “loading custom config file” messages to trace! after the first successful load_and_process_config completion (the flag is set only at the end of that path, so a failed first attempt still logs at info! on retry).

In apollo_config_manager, the periodic poll heartbeat (“periodic check triggered, updating config”) moves from info! to trace!. File-change detection, log_config_diff, and “Successfully updated dynamic config” are unchanged.

^{Reviewed by Cursor Bugbot for commit 7f0051f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[reviewable-StarkWare]

This change is