Fix: remove the leak protection entirely

modules/ANSUtilities/dllmain.cpp

@@ -191,14 +191,21 @@ BOOL APIENTRY DllMain( HMODULE hModule,
     return TRUE;
 }
 
+// Pure constructor: every call allocates a brand-new ANSUtilities and writes it
+// to *Handle. *Handle(in) is ignored completely -- LabVIEW's CLF Node marshalling
+// reuses the same temp buffer per call site, so *Handle(in) often contains
+// leftover bytes from the previous Create's output even when the actual LabVIEW
+// wire is a different, freshly-allocated instance. Inspecting *Handle(in) for
+// "is this live?" detection cannot distinguish that case from a genuine
+// double-Create-on-same-wire, so we don't try -- we trust the caller.
+//
+// Trade-off: if a caller really does call Create twice on the same wire without
+// Release, the first handle leaks. That is the caller's bug. The alternative
+// (releasing live objects we "see" in the input) destroys legitimate parallel
+// instances and is far worse. (Same reasoning as CreateANSAWSHandle.)
 extern "C" ANSULT_API int  CreateANSUtilityHandle(ANSCENTER::ANSUtilities** Handle, const char* licenseKey) {
 	if (Handle == nullptr || licenseKey == nullptr) return 0;
-	if (*Handle) {
-		if (UnregisterUtilHandle(*Handle)) {
-			delete *Handle;
-		}
-		*Handle = nullptr;
-	}
+	*Handle = nullptr;
 	try {
 		*Handle = new ANSCENTER::ANSUtilities();
 		if (*Handle == nullptr) return 0;