RHEL 6.4/6.5 hits an Out of Memory condition (oom-killer is invoked), when there is no memory pressure and more than 8 swap areas are in use

Issue

System has more than 8 swap partitions in use, and frequently experiences an out of memory condition, with oom-killer being invoked. The system may not be under memory pressure, there is still memory available, swap usage is low.

Also, before the oom-killer is invoked, kernel messages similar to this may appear at the kernel log:

swap_free: Unused swap offset entry 0000097f
swap_free: Unused swap offset entry 00000980
BUG: Bad page map in process titan  pte:00130000 pmd:379251a067
addr:00002ac8410e2000 vm_flags:00100077 anon_vma:ffff8836006369c0 mapping:(null) index:2ac8410e2
Pid: 16668, comm: titan Not tainted 2.6.32-431.5.1.el6.x86_64 #1
Call Trace:
 [<ffffffff81145c26>] ? print_bad_pte+0x1d6/0x2b0
 [<ffffffff8115c613>] ? swap_info_get+0xa3/0xe0
 [<ffffffff81148544>] ? unmap_vmas+0x604/0xc50
 [<ffffffff8114e467>] ? exit_mmap+0x87/0x170
 [<ffffffff8106f22c>] ? mmput+0x6c/0x120
 [<ffffffff81190a94>] ? flush_old_exec+0x484/0x690
 [<ffffffff811e45b0>] ? load_elf_binary+0x350/0x1ab0
 [<ffffffff81146222>] ? follow_page+0x412/0x500
 [<ffffffff8114b460>] ? __get_user_pages+0x110/0x430
 [<ffffffff811e185e>] ? load_misc_binary+0x9e/0x3f0
 [<ffffffff8114b819>] ? get_user_pages+0x49/0x50
 [<ffffffff811920a7>] ? search_binary_handler+0x137/0x370
 [<ffffffff81192617>] ? do_execve+0x217/0x2c0
 [<ffffffff810095ea>] ? sys_execve+0x4a/0x80
 [<ffffffff8100b4ca>] ? stub_execve+0x6a/0xc0