Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/actors/benches/anchor_resolution.md`:
- Line 8: The metric name in this document uses the underscored form
irys_mempool_pending_submit_unresolvable_anchors but should use the canonical
dotted form irys.mempool.pending_submit_unresolvable_anchors; update every
occurrence (including the other occurrence referenced) to the dotted form so
queries and dashboards remain consistent, e.g., replace
irys_mempool_pending_submit_unresolvable_anchors with
irys.mempool.pending_submit_unresolvable_anchors wherever it appears.

In `@crates/actors/src/block_tree_service.rs`:
- Around line 99-103: The now_ms function uses an unchecked cast d.as_millis()
as i64; replace it with a checked conversion using i64::try_from (or TryInto) to
avoid truncation/overflow: after obtaining the Duration from
SystemTime::now().duration_since(UNIX_EPOCH) convert d.as_millis() with
i64::try_from(...).ok() (or .and_then(...)) and fall back to 0 on error, keeping
the same unwrap_or(0) semantics; update the conversion in the now_ms function to
use these checked conversions instead of the as cast.

In `@crates/actors/src/mempool_service.rs`:
- Around line 2515-2516: The monitoring interval in mempool_service.rs uses
tokio::time::interval (variable "interval") with the default Burst behavior;
change it to use a safer missed-tick behavior (e.g., MissedTickBehavior::Skip or
::Delay) to avoid rapid catch-up executions — create the interval as before then
call interval.set_missed_tick_behavior(tokio::time::MissedTickBehavior::Skip)
(or ::Delay) before the initial interval.tick().await; reference the "interval"
variable and the tokio::time::MissedTickBehavior enum when applying the change.
- Line 2520: The cast `count as i64` is unsafe for runtime `usize` values;
replace it with a checked conversion using `i64::try_from(count)` and handle the
`Result` (for example, on `Err` cap to `i64::MAX` and emit a warning) before
calling `crate::metrics::set_anchor_unresolvable`; update the usage around
`set_anchor_unresolvable` and the `count` variable to perform `try_from` and a
clear error/overflow handling path instead of the `as` cast.
- Around line 2514-2522: The anchor-resolution sweep task currently spawned via
handle_for_inner.spawn (assigned to sweep_handle) lacks logging; add structured
logs: emit an info/debug at task startup indicating the sweep loop has started
(include interval duration), and emit periodic debug/info logs inside the loop
(e.g., after calling sweep_inner.count_unresolvable_submit_anchors()) that
report the retrieved count and any unexpected errors; reference the spawn block
(sweep_handle), the sweep_inner.count_unresolvable_submit_anchors() call, and
the crate::metrics::set_anchor_unresolvable() call so logs surround the metric
update and surface failures or anomalous counts.

In `@crates/reth-node-bridge/src/node.rs`:
- Around line 267-279: The tests are using NodeConfig::testnet() but should use
the testing constructor; update the test setup to call
irys_types::NodeConfig::testing() (and any ConsensusConfig::testing() if
present) instead of NodeConfig::testnet() so compute_metrics_addr(&cfg, ...)
gets a test-config; locate the occurrences in the tests around
compute_metrics_addr and replace NodeConfig::testnet() with
NodeConfig::testing().

---

Outside diff comments:
In `@crates/p2p/src/block_pool.rs`:
- Around line 769-773: The orphan-retry metric is incremented before the
dispatch, which can falsely count failed sends; move the call to
crate::metrics::record_block_pool_orphan_retry() so it runs only after
self.sync_service_sender.send(...) succeeds. Concretely, in the block handling
SyncChainServiceMessage::AttemptReprocessingBlock(prev_block_hash), perform the
send via self.sync_service_sender.send(...), check the Result (the Err branch
should log the error as before), and call
crate::metrics::record_block_pool_orphan_retry() only in the Ok/success path
(after a successful send) to ensure the metric reflects dispatched retries.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/actors/src/metrics.rs`:
- Around line 133-139: The helpers record_block_tree_last_block_at_ms and
record_block_tree_last_reorg_at_ms (and the other similar helpers around lines
271-273) use runtime `as u64` casts; replace those casts with a checked
conversion using u64::try_from (or TryInto) on the non-negative i64 value and
handle the Result explicitly (e.g., map Err to 0 or log and return) before
calling .record(...), so you avoid unchecked `as` casts in production metric
helpers and update the same pattern in the other functions mentioned.