The last build failed https://ci.chromium.org/p/chrome/builders/luci.chrome.ci/mac-10_13_laptop_high_end-perf/1574
because shard #0 completed with BOT_DIED: https://chrome-swarming.appspot.com/task?id=4096e8f0d7396710

The shard ran for 2h - not sure if it's related, but it seems rather high. Maybe it ran out of memory?

This is not a new failure - another example from Sep 28: https://ci.chromium.org/p/chrome/builders/luci.chrome.ci/mac-10_13_laptop_high_end-perf/1407

The last recent failure of the same kind is in https://ci.chromium.org/p/chrome/builders/luci.chrome.ci/mac-10_13_laptop_high_end-perf/1575 at 2018-10-16 5:22 PM (PDT). The task https://chrome-swarming.appspot.com/user/task/4097a3c17d490710 ran for 140m and failed with BOT_DIED.
The vitals of the bot http://shortn/_QER1j1EcsG seem rather normal - memory at 35%, CPU never exceeded 20% and looks like a regular pattern. There is some discontinuity in a few metrics at 17:00 - right before the build died, but otherwise it seems like a normal reboot.
Something must be killing the swarming bot itself, but I'm not sure what.

I'm not realistically working on this, releasing the bug.