don't instrument long lived tasks 2

[MartinquaXD]

Description

Follow up to #4680 as I overlooked to background tasks.

Changes

adjusted instrumentation of current block stream and token fetcher
also slightly refactored current block stream to reuse duplicated code

[claude]

Claude finished @MartinquaXD's task in 1m 40s —— View job

---

PR Review: don't instrument long lived tasks 2

• Read PR diff and changed files
• Analyze instrumentation changes (block stream + token fetcher)
• Verify refactor preserves original behavior
• Check types/signatures consistency

Summary

Looks good. ✅ The change correctly addresses the goal: long-lived spawned tasks (current_block_stream, current_block_ws_stream, token_fetcher) no longer wrap their whole lifetime in a span that never closes. Instead, #[instrument(skip_all)] is applied to the per-iteration units of work (convert_block_and_process, fetch_block_and_process, update_balances), so spans open and close once per block / per poll / per update cycle.

The refactor of the two block-stream loops into the shared handle_new_block helper is a nice deduplication and I verified it preserves the original control flow exactly:

• update_current_block_metrics still runs for every observed block (before the hash check).
• Same-hash blocks are ignored; update_block_metrics runs only on hash change; the stream is updated + previous_block advanced only when the block number increases.
• The panic-on-send-failure behavior is unchanged (both loops).

Types are consistent (AlloyProvider, get_block_at_id(&AlloyProvider, ...)), and the tracing::Instrument → tracing::instrument import swaps match the new usage in both files.

Minor observations (non-blocking, no change needed):

• The debug log message was unified from "received block" / "polled block" to "observed new block". Since both loops now share code this is expected; the span/task naming still lets you tell WS vs. polling apart.
• BlockInfo::try_from(block.clone()) still clones to retain block for the error log — pre-existing, not introduced here.

I couldn't run cargo check/cargo +nightly fmt in this environment (command approval not granted), so please rely on CI for the final build/format check.
• dont-instrument-long-tasks-2