More robust distributed lock mechanism

[will123195]

We're currently using Redis SETNX to prevent an op from being evented more than once.

Technically during a Redis fail-over, more than one process may receive the lock for an op.

A solution is detailed here:
https://redis.io/commands/setnx

[will123195]

We're already using redis-lock for workers to acquire the "master" worker lock.

We should consider also using it to ensure an op is evented exactly once, although it may come with a performance penalty.