Thanks. But Julliard has pointed out that my fix in patch [2/6] is wrong, and it is. It works, but not for the right reason. I didn't (and still don't) understand how everything in wineserver works together very well. Without the patch [2/6] the conformance tests will fail due to a thread race condition (which works by way of allegory 'by skipping the lock'. there isn't really a lock, but thats the idea). In 'real life' situations the thread race condition is rare, which is why the previous code 'works' for bug 9787.
It's exciting to see such progress on AcceptEx!
But the wine project likes to have all tests passing after each patch in a series. If patch 1/6 adds tests that fail on Wine, they need to be marked with todo_wine. If patch 6/6 fixes the problem in wine that caused those tests to fail, it should also remove the todo_wine from the test.
Following this rule will help make sure your patches don't turn red on http://kegel.com/wine/patchwatcher/results/