diff --git a/clients/rust/crates/mxgw-cli/src/main.rs b/clients/rust/crates/mxgw-cli/src/main.rs
index 21e153b..edcca89 100644
--- a/clients/rust/crates/mxgw-cli/src/main.rs
+++ b/clients/rust/crates/mxgw-cli/src/main.rs
@@ -568,7 +568,7 @@ async fn run(cli: Cli) -> Result<(), Error> {
             json,
         } => {
             let session = session_for(connection, session_id).await?;
-            let results = session.read_bulk(server_handle, items, timeout_ms).await?;
+            let results = session.read_bulk(server_handle, &items, timeout_ms).await?;
             print_read_bulk_results("read-bulk", &results, json);
         }
         Command::WriteBulk {
@@ -1038,7 +1038,7 @@ async fn run_bench_read_bulk(
             + std::time::Duration::from_secs(warmup_seconds);
         while std::time::Instant::now() < warmup_deadline {
             let _ = session
-                .read_bulk(server_handle, tags.clone(), timeout_ms)
+                .read_bulk(server_handle, &tags, timeout_ms)
                 .await;
         }
 
@@ -1052,7 +1052,7 @@ async fn run_bench_read_bulk(
 
         while std::time::Instant::now() < steady_deadline {
             let call_start = std::time::Instant::now();
-            let outcome = session.read_bulk(server_handle, tags.clone(), timeout_ms).await;
+            let outcome = session.read_bulk(server_handle, &tags, timeout_ms).await;
             let elapsed_ms = call_start.elapsed().as_secs_f64() * 1000.0;
             latencies_ms.push(elapsed_ms);
             match outcome {
diff --git a/clients/rust/src/session.rs b/clients/rust/src/session.rs
index dddae21..e7e8930 100644
--- a/clients/rust/src/session.rs
+++ b/clients/rust/src/session.rs
@@ -477,16 +477,28 @@ impl Session {
     /// Per-tag failures appear as `BulkReadResult` entries with
     /// `was_successful = false`; the call never errors on per-tag failure.
     ///
+    /// `tag_addresses` is taken by borrowed slice so callers can re-issue the
+    /// same call repeatedly (typical for the bench loop and for any caller
+    /// polling a fixed snapshot set) without owning or cloning the list at the
+    /// call site. The method still has to materialise an owned `Vec<String>`
+    /// internally because prost's `ReadBulkCommand` field requires it, so this
+    /// is a clarity-of-ownership change rather than an allocation-reducing
+    /// one: total heap traffic per call is unchanged.
+    ///
     /// # Errors
     ///
     /// Same conditions as [`Session::add_item_bulk`].
-    pub async fn read_bulk(
+    pub async fn read_bulk<S: AsRef<str>>(
         &self,
         server_handle: i32,
-        tag_addresses: Vec<String>,
+        tag_addresses: &[S],
         timeout_ms: u32,
     ) -> Result<Vec<BulkReadResult>, Error> {
         ensure_bulk_size("tag_addresses", tag_addresses.len())?;
+        let tag_addresses: Vec<String> = tag_addresses
+            .iter()
+            .map(|s| s.as_ref().to_owned())
+            .collect();
         let reply = self
             .invoke(
                 MxCommandKind::ReadBulk,
diff --git a/clients/rust/tests/client_behavior.rs b/clients/rust/tests/client_behavior.rs
index c156b91..0a50969 100644
--- a/clients/rust/tests/client_behavior.rs
+++ b/clients/rust/tests/client_behavior.rs
@@ -154,7 +154,7 @@ async fn read_bulk_forwards_timeout_and_unpacks_cached_flag() {
     let session = client.session("session-fixture");
 
     let results = session
-        .read_bulk(12, vec!["Area001.Pump001.Speed".to_owned()], 750)
+        .read_bulk(12, &["Area001.Pump001.Speed"], 750)
         .await
         .unwrap();