diff --git a/src/java/org/apache/cassandra/db/guardrails/CMSSizeGuardrail.java b/src/java/org/apache/cassandra/db/guardrails/CMSSizeGuardrail.java
index e1fb4c0b9a..b404d45fcc 100644
--- a/src/java/org/apache/cassandra/db/guardrails/CMSSizeGuardrail.java
+++ b/src/java/org/apache/cassandra/db/guardrails/CMSSizeGuardrail.java
@@ -46,11 +46,8 @@ public class CMSSizeGuardrail extends MinThreshold
 
     public void guard(int totalNodes, int cmsSize)
     {
-        // A cluster can't host more CMS replicas than it has nodes. If the request already uses every node there's
-        // nothing more to enforce (e.g. during bootstrap), and requiring more would brick reconfiguration. Otherwise
-        // the request must meet the configured minimum, even when the threshold exceeds the current cluster size, so
-        // an operator can't sidestep the floor by setting a high threshold. An operator who genuinely wants a smaller
-        // CMS should disable the guardrail. When it is disabled, guard() below is a no-op.
+        // If the request already uses every node, the cluster can't do better, so skip the check.
+        // Otherwise the CMS size must meet the configured minimum.
         if (cmsSize >= totalNodes)
             return;
 
diff --git a/src/java/org/apache/cassandra/db/guardrails/Guardrails.java b/src/java/org/apache/cassandra/db/guardrails/Guardrails.java
index cc44b5add8..7275524a51 100644
--- a/src/java/org/apache/cassandra/db/guardrails/Guardrails.java
+++ b/src/java/org/apache/cassandra/db/guardrails/Guardrails.java
@@ -618,7 +618,8 @@ public final class Guardrails implements GuardrailsMBean
     /**
      * Guardrail on the minimum CMS size (aggregate replication factor across DCs).
      */
-    public static final CMSSizeGuardrail minimumCmsSize = new CMSSizeGuardrail(state -> CONFIG_PROVIDER.getOrCreate(state).getMinimumCmsSizeFailThreshold());
+    public static final CMSSizeGuardrail minimumCmsSize =
+    new CMSSizeGuardrail(state -> CONFIG_PROVIDER.getOrCreate(state).getMinimumCmsSizeFailThreshold());
 
     /**
      * Guardrail on the maximum replication factor.
diff --git a/test/distributed/org/apache/cassandra/distributed/test/guardrails/GuardrailCmsSizeReconfigurationTest.java b/test/distributed/org/apache/cassandra/distributed/test/guardrails/GuardrailCmsSizeReconfigurationTest.java
index e99f15f0ec..ecdf76370d 100644
--- a/test/distributed/org/apache/cassandra/distributed/test/guardrails/GuardrailCmsSizeReconfigurationTest.java
+++ b/test/distributed/org/apache/cassandra/distributed/test/guardrails/GuardrailCmsSizeReconfigurationTest.java
@@ -90,9 +90,8 @@ public class GuardrailCmsSizeReconfigurationTest extends TestBaseImpl
     @Test
     public void reconfigureBelowClusterCapacityIsRejectedWhenThresholdAboveClusterSize() throws IOException
     {
-        // Threshold (4) is above the cluster size (3), but the cluster can still host a CMS of 3. Requesting a smaller
-        // size than the cluster can support must be rejected rather than silently skipped, so a high threshold can't
-        // be used to sidestep the floor.
+        // Threshold (4) is above the cluster size (3), but the cluster can still host a CMS of 3, so a request for 2
+        // is rejected.
         try (Cluster cluster = build(3).withConfig(config(4)).start())
         {
             cluster.get(1).nodetoolResult("cms", "reconfigure", "2")
@@ -105,8 +104,8 @@ public class GuardrailCmsSizeReconfigurationTest extends TestBaseImpl
     @Test
     public void reconfigureAtClusterCapacityIsAllowedWhenThresholdAboveClusterSize() throws IOException
     {
-        // Threshold (4) is above the cluster size (3). A CMS of 3 already uses every node, so it is allowed even though
-        // it is below the threshold: the cluster simply cannot do better, and blocking it would brick reconfiguration.
+        // Threshold (4) is above the cluster size (3). A CMS of 3 already uses every node, so it is allowed even
+        // though it is below the threshold.
         try (Cluster cluster = build(3).withConfig(config(4)).start())
         {
             cluster.get(1).nodetoolResult("cms", "reconfigure", "3")
@@ -115,6 +114,20 @@ public class GuardrailCmsSizeReconfigurationTest extends TestBaseImpl
         }
     }
 
+    @Test
+    public void reconfigureAboveClusterSizeIsNotBlockedByGuardrail() throws IOException
+    {
+        // Requested size (5) exceeds the cluster size (3), so the guardrail does not apply. The reconfiguration still
+        // fails, but because CMS placement can't find enough nodes, not with the guardrail's "failure threshold" error.
+        try (Cluster cluster = build(3).withConfig(config(10)).start())
+        {
+            cluster.get(1).nodetoolResult("cms", "reconfigure", "5")
+                   .asserts()
+                   .failure()
+                   .stderrContains("not enough nodes");
+        }
+    }
+
     @Test
     public void reconfigureMultiDcBelowThresholdIsRejected() throws IOException
     {