Behavior of redundant/load-balance

[Sébastien Barbereau]

Hi
I hope someone will be able to clarify for me the behavior of the
redundant/loadbalance grouping.
In particular, when using the module redundant features (aka failover)
what is the exact way freeradius will proceed?
Let me clarify my question. If I have the following configuration
(from the documentation):
modules {
    sql sql1 {
        # configuration to connect to SQL database one
    }
    sql sql2 {
        # configuration to connect to SQL database two
    }
    always handled {
        rcode = handled
    }
}

#  Handle accounting packets
accounting {
    # always log to detail, stopping if it fails
    detail
    redundant {
    # try module sql1
          sql1
          # if that's down, try module sql2
          sql2
          # otherwise drop the request as
          # it's been "handled" by the "always"
          # module (see doc/rlm_always)
          handled
    }
}

If sql1 is down, freeradius will move to process the request with
sql2, but what does the documentation mean with 'is down' : no
connectivity to DB, slow request, ... ?
Also once it detects that sql1is down, does it "mark" it as down and
afterwards does not query anymore for X minutes?

is there a charitable soul to clarify this ?

Thanks,
Seb.