Resubmission to faulty hardware

[bosterholz]

In some edge cases fine jobs failed because of faulty hardware.
Because of their instant re-submission chances are high, that they are retried on the same faulty instance.
It would be nice to ban instances for certain jobs if they failed there in the past.

Another easier fix would be a retry "sleep" time, in hope that the faulty instance is already booked after the timeout and jobs are scheduled to new instances.

See: Nextflow