A puppet run on db1050 stopped mysqld using an orphan pid file for it own agent, /var/run/puppet/agent.pid, which happened to match the mysqld pid after that service restarted.
150111 15:24:04 [Note] /opt/wmf-mariadb10/bin/mysqld: Normal shutdown
Jan 11 15:24:04 db1050 puppet-agent: (/Stage[main]/Base::Puppet/Service[puppet]/ensure) ensure changed 'running' to 'stopped'
Unsure why this didn't trigger sooner since the orphan pid file timestamp was months old and mysqld had been running for three days.
How to prevent this in the future... sweep check for orphan pid files, or make puppet smarter about blindly stopping services, or treat as a fluke?