Comment 38 for bug 507062

xcb has a history of threading problems, and the bug reported here is probably no exception.

bug #86103 - Locking assertions caused java to totally break. We held off on shipping libxcb-enablement for libx11 for some time due to this bug, until java could be sufficiently fixed to work around it.

bug #87947 - when we finally enabled libxcb in libx11 (we needed it in order to start using Compiz) it spawned a slew of bug reports about locking problems, particularly in (older?) java apps. The main issues were resolved and remaining ones deferred to 185311

bug #185311 - more locking problems. We worked with upstream to resolve many of the asserts and issues that were reported. See comment #156 in particular as my analysis of the issue. Ultimately we "solved" it by switching to sloppy locking.

bug #232364 - A libxcb race condition in dbus plagued several different apps. We figured out we could workaround it by adjusting the order that afflicted apps started things up.

Anyway, I don't know if this history lesson is of much use for the current bug in question, but in case it gives insights or just context, here you go.