I see this once in a while. I am starting up a debug session and open one of our files. The app hangs. take a look at this code:

_XTP_EXT_CLASS CXTPSkinManager* AFX_CDECL XTPSkinManager()
{
    XTP_SKINFRAMEWORK_GLOBALLOCK_SHARED_SCOPE();

    if (CXTPSkinManager::s_pInstance == NULL)
    {
        CXTPSkinManager::s_pInstance = &XTPSkinManagerSingleton::Instance();
    }

    return CXTPSkinManager::s_pInstance;
}

Note this line:

    XTP_SKINFRAMEWORK_GLOBALLOCK_SHARED_SCOPE();

On the main thread the stack is:

     ToolkitPro2410vc170x64UD.dll!CXTPSimpleCriticalSection::Enter() Line 100    C++
     ToolkitPro2410vc170x64UD.dll!CXTPSimpleCriticalSection::CLock::CLock(CXTPSimpleCriticalSection & cs) Line 169    C++
     ToolkitPro2410vc170x64UD.dll!CXTPRWCriticalSection::LockShared(int bPopMode) Line 327    C++
     ToolkitPro2410vc170x64UD.dll!CXTPRWCriticalSection::LockShared() Line 227    C++
     ToolkitPro2410vc170x64UD.dll!CXTPRWCriticalSection::CSharedLock::CSharedLock(CXTPRWCriticalSection & cs) Line 311    C++
>    ToolkitPro2410vc170x64UD.dll!XTPSkinManager() Line 914    C++

In the Enter() function I am stuck in this loop:

            while (0 != m_nLock)
            {
                Sleep(dwTimeout);
                if (dwTimeout < XTP_SYNC_MAX_EXPONENTIAL_TIMEOUT)
                {
                    dwTimeout <<= 1;
                }
            }

Examining the plethora of ToolkitPro worker threads I find this one:

Notice that thread is also calling:

XTP_EXT_CLASS CXTPSkinManager* AFX_CDECL XTPSkinManager()
{
    XTP_SKINFRAMEWORK_GLOBALLOCK_SHARED_SCOPE();

    if (CXTPSkinManager::s_pInstance == NULL)
    {
        CXTPSkinManager::s_pInstance = &XTPSkinManagerSingleton::Instance();
    }

    return CXTPSkinManager::s_pInstance;
}

Setting breakpoints I see the main thread is in the loop. Meanwhile the second thread never returns from the RtlpWaitOnCriticalObject.

Since we have entered idle processing and we are a COM app, the OS is freeing unused libraries. That goes thru _AfxCrtDumpClient which is what the main thread is hooking into when memory leak detection is running and the call to get the mscoree.dll is made:

void __cdecl _AfxCrtDumpClient(void * pvData, size_t nBytes)
{
    if(_CrtReportBlockType(pvData) != _AFX_CLIENT_BLOCK)
        return;
    char sz[256];
    CObject* pObject = (CObject*)pvData;
    bool fCLRPresent=(::GetModuleHandleW(L"mscoree.dll")!=NULL);

All other ToolkitPro threads look standard and are all "small stacks" waiting on an object.

When this happens all I can do is restart.