diff --git a/packages/services/pegboard/src/workflows/datacenter.rs b/packages/services/pegboard/src/workflows/datacenter.rs
index 0f0dd09567..9ff9e6ba82 100644
--- a/packages/services/pegboard/src/workflows/datacenter.rs
+++ b/packages/services/pegboard/src/workflows/datacenter.rs
@@ -124,7 +124,6 @@ async fn allocate_actor(
 		"allocating actor"
 	);
 
-	// TODO(RVT-4455): Validate this isolate vs container logic
 	// Even though isolates autoscale based on CPU, we allocate machines based on reservation in
 	// balance proactively. Otherwise, we'd end up with bad scaling with retroactively choosing
 	// nodes based on CPU load since actors will show the CPU load after a delay.
@@ -173,23 +172,25 @@ async fn allocate_actor(
 		SELECT $3, client_id, $4, $5
 		FROM available_clients
 		WHERE
-			-- Containers: ensure has available resources
-			-- Isolates: don't limit resources for isolates since they scale on CPU
-			$8 = 1 OR
-			(
-				allocated_cpu + $6 <= available_cpu AND
-				allocated_memory + $7 <= available_memory
-			)
-		-- Container: binpack to the most-populated node to maximize density
-		-- Isolate: allocate to the least-allocated machine since we autoscale on CPU
+			-- Containers (0): ensure node has available resources
+			-- Isolates (1): don't limit resources since they scale on CPU
+			CASE WHEN $8 = 0
+				THEN (
+					allocated_cpu + $6 <= available_cpu AND
+					allocated_memory + $7 <= available_memory
+				)
+				ELSE TRUE
+			END
 		ORDER BY
-			CASE
-				WHEN $8 = 0 THEN -allocated_cpu
-				ELSE available_cpu
-			END ASC,
-			CASE
-				WHEN $8 = 0 THEN -allocated_memory
-				ELSE available_memory
+			-- Container (0): binpack to the most-populated node to maximize density
+			-- Isolate (1): allocate to the least-populated node since we autoscale on CPU
+			CASE WHEN $8 = 0
+				THEN allocated_cpu
+				ELSE -allocated_cpu
+			END DESC,
+			CASE WHEN $8 = 0
+				THEN allocated_memory
+				ELSE -allocated_memory
 			END DESC
 		LIMIT 1
 		RETURNING client_id