https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91115
--- Comment #6 from Fred Hsueh <fhsueh at roku dot com> --- Sorry for the long silence ... more on this: I tracked down where the std::thread() that was causing this issue. It seems that there are many similar long running tasks that needs to be completed so we keep a queue of 10x threads. As one finishes, another is kicked off. But initially, 10x such threads are started in quick succession. This leads me to believe maybe there's a gap in the locking on the ASAN metadata side. I'll see if I can get an example going.