Hi Stephen, Can you attach to the stuck nm process from context 1 using gdb and then look at the backtrace to narrow down which code is doing this? This might not tell you much if the lock is used from python code, but nonetheless might provide interesting information. Marc