As a user of the master-slave system I think this enhancement is something that
will make hudson practical for scaling up and is really needed thats why I'm
putting it a P1.

Right now the hudson instance sometimes loses slaves - I'm not sure why it
happens, but hey it loses them sometimes. But the thing is I need to manually
launch the slaves even if I've configured them to be started via SSH.

This is really tiresome and hudson should try to revive slaves automatically -
specially if jobs have been queued that require the slave. Having to go there
and manually start slaves is really unacceptable specially when you scale up and
want automation.