Schedule regular production job restarts

[melange396]

Anecdotal instances suggest that such restarts can cause performance increases (or at least ameliorate accumulated performance problems). Before committing to scheduling this in production, we should do some experimentation (probably in production) to verify and measure the performance differences.

Look into regularly scheduling restarts of long-running production jobs, like mysqld and our web app containers. Something like a "round robin" approach might help minimize any impact on users.