arm devices on Chrome OS are posting roughly 70-90 renderer crashes per million page views at the moment - roughly 10x what we'd normally expect to see.

https://crash.corp.google.com/browse?q=product.Version%3D%2757.0.2987.75%27%20AND%20cpu.architecture%3D%27arm%27%20AND%20product.name%3D%27Chrome_ChromeOS%27%20AND%20custom_data.ChromeCrashProto.ptype%3D%27renderer%27&ignore_case=false&enable_rewrite=true&omit_field_name=&omit_field_value=&omit_field_opt=%3D has a breakdown of the crashes for the latest beta build.

60+% of them are hosed stacks that map to __clock_gettime.
The next two buckets are TCMalloc Corruption and base:TimeTicks:Now, also with corrupted stacks. Collectively over 80% of crashes have garbage stacks.

I suspect there are actually multiple issues getting lumped into the same bucket, but having that many crashes dumped into opaque buckets makes stability work on arm next to impossible.

jamescook@ added a comment to __clock_gettime that that could be caused by people working on crash reporting. CCing for context.