Ive just looked to an interesting blog article talking about when a process terminates. http://blog.kalmbachnet.de/?postid=65 It seem that there is a subtle difference between how a console application (the main() CRT function) and a Win32 application (WinMain) is handling the process termination. if you have this: DWORD WINAPI MyThread2(LPVOID) { while(true) { } return 0; } #pragma comment(linker, "/entry:myMain") int WINAPI myMain() { DWORD dwThreadId; CloseHandle(CreateThread(NULL, 0, MyThread2, NULL, 0, &dwThreadId)); return 0; } the application is a windows application (using myMain entrypoint) and it run infinitely. but a similar console application: int t_main() { DWORD dwThreadId; CloseHandle(CreateThread(NULL, 0, MyThread2, NULL, 0, &dwThreadId)); return 0; } will return immediately after the 'return 0;' the difference is this one: the console application uses CRT which, after the user code return from main, it calls ExitProcess which terminates the a...

as the msdn says SuspendThread is for debugging purposes and should not be used. it may be tempting to use it to pause a worker thread for example but it might not be ok. the worker thread can be in the middle of executing a CRT function which is thread safe, using sychronization. if the worker thread is paused after the lock is done, trying to execute from another thread a CRT function, which uses the same syncronization access, might get to a deadlock.