As far as I can tell, no other thread is active.  Their backtraces are all either:


#0  0x00007f283dedd300 in sem_wait () from /lib/x86_64-linux-gnu/libpthread.so.0
#1  0x0000000000519295 in PyThread_acquire_lock (lock=0xe7dd50, waitflag=1) at Python/thread_pthread.h:321
#2  0x00000000004c734f in PyEval_RestoreThread (tstate=0xfbfb40) at Python/ceval.c:357


or

Thread 77 (Thread 0x7f2808fb1700 (LWP 30292)):
#0  0x00007f283d57a773 in poll () from /lib/x86_64-linux-gnu/libc.so.6
#1  0x00007f283c458739 in internal_select (s=0x7f2831432ca0, writing=0) at /root/python2.7-2.7.2/Modules/socketmodule.c:692
#2  0x00007f283c45acdf in sock_recv_guts (s=0x7f2831432ca0, cbuf=0x339bf7c "\313\313\313\313\313\313\313\313\313\313\313\313\313\313\313", <incomplete sequence \313>, len=16, flags=0) at /root/python2.7-2.7.2/Modules/socketmodule.c:2356
(More stack frames follow...)

or

Thread 17 (Thread 0x7f2829ff3700 (LWP 19252)):
#0  0x00007f283d57f913 in select () from /lib/x86_64-linux-gnu/libc.so.6
#1  0x00007f283d2a1fac in floatsleep (secs=0) at /root/python2.7-2.7.2/Modules/timemodule.c:943
#2  0x00007f283d2a0cd9 in time_sleep (self=0x0, args=(<float at remote 0xb3ed40>,)) at /root/python2.7-2.7.2/Modules/timemodule.c:206
(More stack frames follow...)

or
Thread 12 (Thread 0x7f2834e5f700 (LWP 19223)):
#0  0x00007f283d57f913 in select () from /lib/x86_64-linux-gnu/libc.so.6
#1  0x00007f2839cb275e in select_select (self=0x0, args=([<_socketobject at remote 0x7f2830f4d1a0>, <_socketobject at remote 0x2106ec0>], [], [], <float at remote 0x7f283034d510>)) at /root/python2.7-2.7.2/Modules/selectmodule.c:278
#2  0x000000000055f0ee in PyCFunction_Call (func=<built-in function select>, arg=([<_socketobject at remote 0x7f2830f4d1a0>, <_socketobject at remote 0x2106ec0>], [], [], <float at remote 0x7f283034d510>), kw=0x0) at Objects/methodobject.c:81
(More stack frames follow...)


Unfortunately, I've only been able to produce the segfault with this program (an EC2 server monitor and deployer).  It's very site specific and couldn't run on another machine.

That said, I can trigger these segfaults with high reliability, so I can certainly add additional debugging logs if requested.