MSDP RPF question

The RPF rules of draft-13 don't well specify how to handle the case
that one of the rules (i) ... (v) returns a P with which we are
MSDP peering, but for which the MSDP peering is not up and running.
Can someone explain to me how the authors of the spec intended that
situation to be handled ?

I for once can see that if such a case happens in rule (ii), that 
i actually may want to fall back checking for rule (iii) and further,
or at least rule (v), right ? After all, the RPF check rules should
cope with the fact that the "best" MSDP peer is not available (even 
though the MRIB doesn't have a clue).

On the other hand, if the same thing happens in rule (i), then should
i not always want to return RP anyhow ? For once, there is the argument,
that a direct MSDP connection to RP should be a good indication as to
wether or not RP is actually running and supposed to generate SAs. In
addition, i can imagine cases in redundant topologies where one might
otherwise create loops for RPs SAs in the case where direct MSDP peering
to RP go away and the RPF check falls back to other rules (not particular
good topologies, but somehow necessary ones).

So, how is the RPF check supposed to work with respect to MSDP
peerings being up and running ? Did i miss to read the paragraph that
explicitly specifies how ?

Cheers
	Toerless