diff --git a/hive-ag3nt/src/stats.rs b/hive-ag3nt/src/stats.rs
index ff28d916..020dec90 100644
--- a/hive-ag3nt/src/stats.rs
+++ b/hive-ag3nt/src/stats.rs
@@ -367,9 +367,15 @@ fn fill_buckets(
             let avg = if sorted.is_empty() {
                 0.0
             } else {
-                #[allow(clippy::cast_precision_loss)]
+                #[allow(
+                    clippy::cast_precision_loss,
+                    reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+                )]
                 let sum_f = sorted.iter().sum::<i64>() as f64;
-                #[allow(clippy::cast_precision_loss)]
+                #[allow(
+                    clippy::cast_precision_loss,
+                    reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+                )]
                 let len_f = sorted.len() as f64;
                 sum_f / len_f
             };
@@ -378,9 +384,15 @@ fn fill_buckets(
             let avg_ctx = if acc.turn_count == 0 {
                 0.0
             } else {
-                #[allow(clippy::cast_precision_loss)]
+                #[allow(
+                    clippy::cast_precision_loss,
+                    reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+                )]
                 let sum_f = acc.ctx_sum as f64;
-                #[allow(clippy::cast_precision_loss)]
+                #[allow(
+                    clippy::cast_precision_loss,
+                    reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+                )]
                 let cnt_f = acc.turn_count as f64;
                 sum_f / cnt_f
             };
@@ -427,9 +439,15 @@ fn summarize_durations(all: &mut [i64]) -> DurationSummary {
         return DurationSummary::default();
     }
     all.sort_unstable();
-    #[allow(clippy::cast_precision_loss)]
+    #[allow(
+        clippy::cast_precision_loss,
+        reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+    )]
     let sum_f = all.iter().sum::<i64>() as f64;
-    #[allow(clippy::cast_precision_loss)]
+    #[allow(
+        clippy::cast_precision_loss,
+        reason = "stat magnitudes (counts, token + duration sums) stay well under f64's 2^53 exact-integer range, so this averaging cast loses no precision in practice"
+    )]
     let len_f = all.len() as f64;
     DurationSummary {
         avg_ms: sum_f / len_f,
diff --git a/hive-c0re/src/container_stats.rs b/hive-c0re/src/container_stats.rs
index ea488da4..bd39097e 100644
--- a/hive-c0re/src/container_stats.rs
+++ b/hive-c0re/src/container_stats.rs
@@ -141,16 +141,25 @@ pub async fn gather() -> Vec<ContainerResource> {
     let t0: Vec<Option<u64>> = candidates.iter().map(|(_, d)| read_usage_usec(d)).collect();
     sleep(CPU_SAMPLE).await;
 
-    #[allow(clippy::cast_precision_loss)]
+    #[allow(
+        clippy::cast_precision_loss,
+        reason = "CPU-utilisation math; the operands (core count, microsecond sample interval, cgroup time delta) stay well under f64's 2^53 exact-integer range, so no precision is lost"
+    )]
     let nproc = host_nproc() as f64;
-    #[allow(clippy::cast_precision_loss)]
+    #[allow(
+        clippy::cast_precision_loss,
+        reason = "CPU-utilisation math; the operands (core count, microsecond sample interval, cgroup time delta) stay well under f64's 2^53 exact-integer range, so no precision is lost"
+    )]
     let interval_usec = CPU_SAMPLE.as_micros() as f64;
 
     let mut out: Vec<ContainerResource> = Vec::with_capacity(candidates.len());
     for (i, (name, dir)) in candidates.iter().enumerate() {
         let cpu_pct = match (t0[i], read_usage_usec(dir)) {
             (Some(a), Some(b)) => {
-                #[allow(clippy::cast_precision_loss)]
+                #[allow(
+                    clippy::cast_precision_loss,
+                    reason = "CPU-utilisation math; the operands (core count, microsecond sample interval, cgroup time delta) stay well under f64's 2^53 exact-integer range, so no precision is lost"
+                )]
                 let delta = b.saturating_sub(a) as f64;
                 (delta / (interval_usec * nproc)) * 100.0
             }
diff --git a/hive-c0re/src/hive_stats.rs b/hive-c0re/src/hive_stats.rs
index 28e3d44b..e70f984d 100644
--- a/hive-c0re/src/hive_stats.rs
+++ b/hive-c0re/src/hive_stats.rs
@@ -205,7 +205,10 @@ fn read_agent(path: &Path, from: i64) -> rusqlite::Result<AgentAgg> {
         agg.cache_read = agg.cache_read.saturating_add(cache_read);
         agg.cache_creation = agg.cache_creation.saturating_add(cache_creation);
         let p = model_prices(&model);
-        #[allow(clippy::cast_precision_loss)]
+        #[allow(
+            clippy::cast_precision_loss,
+            reason = "token counts stay well under f64's 2^53 exact-integer range, so this cost computation loses no precision"
+        )]
         {
             agg.cost += (input as f64 * p.input
                 + output as f64 * p.output