SQS Message handling settings don't work.

[alexgleith]

It should be possible to use Statistician configuration parameters for crop mask like this:

          # SQS Message Handling
          max_processing_time: 0    # Take as long as you want
          job_queue_max_lease: 600  # Lease for 10 minutes at a time
          renew_safety_margin: 30   # Extend with 30 seconds remaining
          future_poll_interval: 10  # Check every 10 seconds

But this doesn't work, as evidenced by the weird behaviour in messages below.

What I think is happening is that messages are going back on the queue and being retried, so that they end up being run multiple times in parallel... this should not happen as it's extremely wasteful of resources.

The solution is to use long timeouts, which slows down retries for real failures.