Add a DB "ping" for each worker

Each worker should be able to connect through the load balancer and perform a "ping" to ensure connectivity every minute or so. If this ever fails for N consecutive tries, we should know ASAP.

This applies both to Redis and PostgreSQL.