After JDK-8315818, it's possible for JVMCI initialization to happen prior to the forced initialization in Threads::create_vm. The former will happen on a compiler thread where as the latter happens on the main thread. This has the unfortunate side effect of JVMCI property printing happening twice.