diff --git a/public/client/TracyThread.hpp b/public/client/TracyThread.hpp
index f95627d4..f22d6c5e 100644
--- a/public/client/TracyThread.hpp
+++ b/public/client/TracyThread.hpp
@@ -42,22 +42,56 @@ public:
     Thread( void(*func)( void* ptr ), void* ptr )
         : m_func( func )
         , m_ptr( ptr )
+        , m_exit( CreateEventW( nullptr, TRUE, FALSE, nullptr ) )
         , m_hnd( CreateThread( nullptr, 0, Launch, this, 0, nullptr ) )
     {}
 
     ~Thread()
     {
-        WaitForSingleObject( m_hnd, INFINITE );
+        // The thread signals m_exit when it is done and then parks, instead of exiting, so that
+        // it can be joined from DllMain (e.g. a static Profiler destroyed by FreeLibrary).
+        // A normal thread exit needs the loader lock, which the joining thread already holds.
+        HANDLE handles[2] = { m_exit, m_hnd };
+        WaitForMultipleObjects( 2, handles, FALSE, INFINITE );
+        CloseHandle( m_exit );
+        if( InLoaderCallout() )
+        {
+            TerminateThread( m_hnd, 0 );
+        }
+        else
+        {
+            QueueUserAPC( WakeUp, m_hnd, 0 );
+            WaitForSingleObject( m_hnd, INFINITE );
+        }
         CloseHandle( m_hnd );
     }
 
     HANDLE Handle() const { return m_hnd; }
 
 private:
-    static DWORD WINAPI Launch( void* ptr ) { ((Thread*)ptr)->m_func( ((Thread*)ptr)->m_ptr ); return 0; }
+    static bool InLoaderCallout()
+    {
+        typedef BOOLEAN(NTAPI* IsWithinCallout)();
+        static const auto fn = (IsWithinCallout)GetProcAddress( GetModuleHandleW( L"ntdll.dll" ), "RtlIsThreadWithinLoaderCallout" );
+        return fn && fn();
+    }
+
+    static void CALLBACK WakeUp( ULONG_PTR ) {}
+
+    static DWORD WINAPI Launch( void* ptr )
+    {
+        auto self = (Thread*)ptr;
+        self->m_func( self->m_ptr );
+        // *self may be gone as soon as m_exit is signaled.
+        const HANDLE exit = self->m_exit;
+        SetEvent( exit );
+        SleepEx( INFINITE, TRUE );
+        return 0;
+    }
 
     void(*m_func)( void* ptr );
     void* m_ptr;
+    HANDLE m_exit;
     HANDLE m_hnd;
 };