Since this seems to be related to the optimization level on my system, it could be a result of some optimization in our GPU code, and the nvidia driver on the stack is a red herring.

There's a comment in stack_trace_posix.cc that seems to imply we need to warm up backtrace(), otherwise it will hang with a very similar looking trace.

https://cs.chromium.org/chromium/src/base/debug/stack_trace_posix.cc?type=cs&q=StackTrace::StackTrace&sq=package:chromium&l=440
"""
  // Warm up stack trace infrastructure. It turns out that on the first
  // call glibc initializes some internal data structures using pthread_once,
  // and even backtrace() can call malloc(), leading to hangs.
  //
  // Example stack trace snippet (with tcmalloc):
  //
  // #8  0x0000000000a173b5 in tc_malloc
  //             at ./third_party/tcmalloc/chromium/src/debugallocation.cc:1161
  // #9  0x00007ffff7de7900 in _dl_map_object_deps at dl-deps.c:517
  // #10 0x00007ffff7ded8a9 in dl_open_worker at dl-open.c:262
  // #11 0x00007ffff7de9176 in _dl_catch_error at dl-error.c:178
  // #12 0x00007ffff7ded31a in _dl_open (file=0x7ffff625e298 "libgcc_s.so.1")
  //             at dl-open.c:639
  // #13 0x00007ffff6215602 in do_dlopen at dl-libc.c:89
  // #14 0x00007ffff7de9176 in _dl_catch_error at dl-error.c:178
  // #15 0x00007ffff62156c4 in dlerror_run at dl-libc.c:48
  // #16 __GI___libc_dlopen_mode at dl-libc.c:165
  // #17 0x00007ffff61ef8f5 in init
  //             at ../sysdeps/x86_64/../ia64/backtrace.c:53
  // #18 0x00007ffff6aad400 in pthread_once
  //             at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_once.S:104
  // #19 0x00007ffff61efa14 in __GI___backtrace
  //             at ../sysdeps/x86_64/../ia64/backtrace.c:104
  // #20 0x0000000000752a54 in base::debug::StackTrace::StackTrace
  //             at base/debug/stack_trace_posix.cc:175
  // #21 0x00000000007a4ae5 in
  //             base::(anonymous namespace)::StackDumpSignalHandler
  //             at base/process_util_posix.cc:172
  // #22 <signal handler called>
"""

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/193e9210f91a8fc07f26d405cb1be69183060e9d

commit 193e9210f91a8fc07f26d405cb1be69183060e9d
Author: Brett Wilson <brettw@chromium.org>
Date: Wed Sep 13 16:09:10 2017

Initialize stack dumping in OOP memlog.

Failing to explicitly initialize stack dumping before adding a malloc hook that
uses it can cause problems. In particular, on Linux calling backtrace() can
load a library which calls malloc, meaning the hook will be recursive.

Bug:  764522 
Change-Id: If602902aa4238903541ca52f765927369dbf6e71
Reviewed-on: https://chromium-review.googlesource.com/664301
Reviewed-by: Erik Chen <erikchen@chromium.org>
Commit-Queue: Erik Chen <erikchen@chromium.org>
Cr-Commit-Position: refs/heads/master@{#501644}
[modify] https://crrev.com/193e9210f91a8fc07f26d405cb1be69183060e9d/chrome/common/profiling/memlog_allocator_shim.cc