erosion resistance.

[Filirom1]

If an instances exits/errors it should schedule a new job to bring it up again on a dynohost

[Filirom1]

In updatestate handle logic to handle failed instance, retries, and kills - if dyno is not consistent with state in database.

[Filirom1]

The following commit b0b784a 2d1bd23 91a98f4 improve the situation, when the ps api was not consistent