Waiting for a worker should timeout or failed

[fsamin]

with only binary requirements, if no worker model matches the requirements; the job should failed as fast as possible. if the requirements is already known by CDS: we can failed when the job if starting if the requirement is unknown we have to wait for all worker registrations

[sguiheux]

It seems that « no hatcheries » match your requirement is not working anymore. Moreover, when this happen, job must fail