diff --git a/dockerutil/fileretriever.go b/dockerutil/fileretriever.go
index 46f8729a1..35d486b10 100644
--- a/dockerutil/fileretriever.go
+++ b/dockerutil/fileretriever.go
@@ -51,8 +51,12 @@ func (r *FileRetriever) SingleFileContent(ctx context.Context, volumeName, relPa
 			Labels: map[string]string{CleanupLabel: r.testName},
 		},
 		&container.HostConfig{
-			Binds:      []string{volumeName + ":" + mountPath},
-			AutoRemove: true,
+			// No AutoRemove: this container is never started (we only
+			// CopyFromContainer its filesystem), and the deferred
+			// ContainerRemove(Force) below cleans it up. Keeping the one-off
+			// helpers uniformly AutoRemove-free avoids the ContainerWait race
+			// that AutoRemove causes on newer Docker Engines (see volumeowner.go).
+			Binds: []string{volumeName + ":" + mountPath},
 		},
 		nil, // No networking necessary.
 		nil,
diff --git a/dockerutil/filewriter.go b/dockerutil/filewriter.go
index 3db357825..8cfbbfd74 100644
--- a/dockerutil/filewriter.go
+++ b/dockerutil/filewriter.go
@@ -59,8 +59,7 @@ func (w *FileWriter) WriteFile(ctx context.Context, volumeName, relPath string,
 			Labels: map[string]string{CleanupLabel: w.testName},
 		},
 		&container.HostConfig{
-			Binds:      []string{volumeName + ":" + mountPath},
-			AutoRemove: true,
+			Binds: []string{volumeName + ":" + mountPath},
 		},
 		nil, // No networking necessary.
 		nil,
@@ -70,13 +69,12 @@ func (w *FileWriter) WriteFile(ctx context.Context, volumeName, relPath string,
 		return fmt.Errorf("creating container: %w", err)
 	}
 
-	autoRemoved := false
+	// We intentionally do NOT set AutoRemove on the container. With AutoRemove
+	// the daemon reaps the container the instant its (near-instant) chown exits,
+	// which on newer Docker Engines (25+) races the ContainerWait below and
+	// surfaces as "No such container". Instead we let the container persist and
+	// always remove it ourselves in this defer.
 	defer func() {
-		if autoRemoved {
-			// No need to attempt removing the container if we successfully started and waited for it to complete.
-			return
-		}
-
 		if err := w.cli.ContainerRemove(ctx, cc.ID, types.ContainerRemoveOptions{
 			Force: true,
 		}); err != nil {
@@ -127,8 +125,6 @@ func (w *FileWriter) WriteFile(ctx context.Context, volumeName, relPath string,
 	case err := <-errCh:
 		return err
 	case res := <-waitCh:
-		autoRemoved = true
-
 		if res.Error != nil {
 			return fmt.Errorf("waiting for write-file container: %s", res.Error.Message)
 		}
diff --git a/dockerutil/volumeowner.go b/dockerutil/volumeowner.go
index 8432ab2a6..f3fd12f38 100644
--- a/dockerutil/volumeowner.go
+++ b/dockerutil/volumeowner.go
@@ -58,8 +58,7 @@ func SetVolumeOwner(ctx context.Context, opts VolumeOwnerOptions) error {
 			Labels: map[string]string{CleanupLabel: opts.TestName},
 		},
 		&container.HostConfig{
-			Binds:      []string{opts.VolumeName + ":" + mountPath},
-			AutoRemove: true,
+			Binds: []string{opts.VolumeName + ":" + mountPath},
 		},
 		nil, // No networking necessary.
 		nil,
@@ -69,13 +68,12 @@ func SetVolumeOwner(ctx context.Context, opts VolumeOwnerOptions) error {
 		return fmt.Errorf("creating container: %w", err)
 	}
 
-	autoRemoved := false
+	// We intentionally do NOT set AutoRemove on the container. With AutoRemove
+	// the daemon reaps the container the instant its (near-instant) chown exits,
+	// which on newer Docker Engines (25+) races the ContainerWait below and
+	// surfaces as "No such container". Instead we let the container persist and
+	// always remove it ourselves in this defer.
 	defer func() {
-		if autoRemoved {
-			// No need to attempt removing the container if we successfully started and waited for it to complete.
-			return
-		}
-
 		if err := opts.Client.ContainerRemove(ctx, cc.ID, types.ContainerRemoveOptions{
 			Force: true,
 		}); err != nil {
@@ -94,8 +92,6 @@ func SetVolumeOwner(ctx context.Context, opts VolumeOwnerOptions) error {
 	case err := <-errCh:
 		return err
 	case res := <-waitCh:
-		autoRemoved = true
-
 		if res.Error != nil {
 			return fmt.Errorf("waiting for volume-owner container: %s", res.Error.Message)
 		}