diff --git a/packages/api/internal/orchestrator/delete_instance.go b/packages/api/internal/orchestrator/delete_instance.go
index 5d1b979bcd..1528c386e7 100644
--- a/packages/api/internal/orchestrator/delete_instance.go
+++ b/packages/api/internal/orchestrator/delete_instance.go
@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"time"
 
 	"github.com/google/uuid"
 	"go.uber.org/zap"
@@ -80,11 +81,17 @@ func (o *Orchestrator) RemoveSandbox(ctx context.Context, teamID uuid.UUID, sand
 	if alreadyDone {
 		logger.L().Info(ctx, "Sandbox was already in the process of being removed", logger.WithSandboxID(sandboxID), zap.String("state", string(sbx.State)))
 
+		if time.Since(sbx.EndTime) > sandbox.StaleCutoff && opts.Action.Effect == sandbox.TransitionExpires {
+			o.sandboxStore.Remove(context.WithoutCancel(ctx), teamID, sandboxID)
+			go o.analyticsRemove(context.WithoutCancel(ctx), sbx, opts.Action)
+		}
+
 		return nil
 	}
 
 	defer func() { go o.analyticsRemove(context.WithoutCancel(ctx), sbx, opts.Action) }()
-	defer o.sandboxStore.Remove(ctx, teamID, sandboxID)
+	// Once we start the removal process, we want to make sure it gets removed from the store
+	defer o.sandboxStore.Remove(context.WithoutCancel(ctx), teamID, sandboxID)
 	err = o.removeSandboxFromNode(ctx, sbx, opts.Action)
 	if err != nil {
 		logger.L().Error(ctx, "Error pausing sandbox", zap.Error(err), logger.WithSandboxID(sbx.SandboxID))
diff --git a/packages/api/internal/sandbox/states.go b/packages/api/internal/sandbox/states.go
index 0a59155310..aa267ed9db 100644
--- a/packages/api/internal/sandbox/states.go
+++ b/packages/api/internal/sandbox/states.go
@@ -4,6 +4,10 @@ import (
 	"time"
 )
 
+// StaleCutoff is how long a team entry must be expired before it can be pruned
+// This prevents races where TTL is extended, but the sandbox would still be removed
+const StaleCutoff = 10 * time.Minute
+
 // TransitionEffect describes what happens to the sandbox when a state action completes.
 type TransitionEffect int
 
diff --git a/packages/api/internal/sandbox/storage/redis/items.go b/packages/api/internal/sandbox/storage/redis/items.go
index 77c8b860d8..6539aa8aa9 100644
--- a/packages/api/internal/sandbox/storage/redis/items.go
+++ b/packages/api/internal/sandbox/storage/redis/items.go
@@ -114,7 +114,8 @@ func (s *Storage) ExpiredItems(ctx context.Context) ([]sandbox.Sandbox, error) {
 
 			// Only evict running sandboxes
 			if sbx.State != sandbox.StateRunning {
-				if time.Since(sbx.EndTime) <= staleCutoff {
+				// If the sandbox is in transitioning state for more than stale cutoff, it's likely failed removal. Let it be cleaned up by the regular expiration process.
+				if time.Since(sbx.EndTime) <= sandbox.StaleCutoff {
 					// Let the current removal finish
 
 					continue
diff --git a/packages/api/internal/sandbox/storage/redis/operations.go b/packages/api/internal/sandbox/storage/redis/operations.go
index f67db89ee0..87ae8ae790 100644
--- a/packages/api/internal/sandbox/storage/redis/operations.go
+++ b/packages/api/internal/sandbox/storage/redis/operations.go
@@ -224,11 +224,6 @@ func (s *Storage) Update(ctx context.Context, teamID uuid.UUID, sandboxID string
 	return updatedSbx, nil
 }
 
-// staleCutoff is how long a team entry must be idle (no Add calls) before it
-// can be pruned from the global teams ZSET when its sandbox count is zero.
-// This prevents races where a Remove sees SCARD==0 right before an Add.
-const staleCutoff = time.Hour
-
 func (s *Storage) TeamsWithSandboxCount(ctx context.Context) (map[uuid.UUID]int64, error) {
 	members, err := s.redisClient.ZRangeWithScores(ctx, globalTeamsSet, 0, -1).Result()
 	if err != nil {
@@ -269,7 +264,7 @@ func (s *Storage) TeamsWithSandboxCount(ctx context.Context) (map[uuid.UUID]int6
 	}
 
 	now := time.Now().Unix()
-	cutoff := now - int64(staleCutoff.Seconds())
+	cutoff := now - int64(sandbox.StaleCutoff.Seconds())
 
 	teams := make(map[uuid.UUID]int64, len(entries))
 	var stale []any