some pipeline jobs hangs forever without hitting timeout

[gberche-orange]

Currently, we sometimes observe some pipelines whose jobs hangs forever. We currently have to manually cancel these hang jobs. This seems similar to https://github.com/concourse/concourse/issues/456

The prometheus build duration is'nt yet helping (see https://github.com/bosh-prometheus/prometheus-boshrelease/issues/173 ) to alert such conditions.

According to http://concourse-ci.org/timeout-step.html each task can have timeouts on any steps (get, tasks, do, ...) See related https://github.com/concourse/concourse/issues/757