caddy-frontend: Raise promise only on slave errors impacting cluster

If slaves clashes domain, this can lead to very bad situation, then only in
this case raise a promise error.

Otherwise the slave requester is informed anyway about the problem, so there
is no reason to make the whole cluster raising an monitoring issue.
10 jobs for master
in 0 seconds, using 0 compute credits, and was queued for 0 seconds