After discussing with lars in the meeting recently and having done some testing, let's: * not use envconsul at all for collector * write crashes to disk * run crashmover (using envconsul) to move crashes from disk->S3 Otherwise, collector will not start up if the consul server nodes are down/unreachable. I think for other services this risk is acceptable, but If we can figure out another way to get the configuration to the collector in the case of consul being down, we could have collector first try writing to S3 and fall back to disk (maybe using hiera-S3? we already depend on S3 being up to provision our collector ASG nodes.) I think we'd just need to provide the name of the bucket to store crashes in.

PR for running crashmover: https://github.com/mozilla/socorro-infra/pull/47