consistent policy != consistent announcements

I think that his internal routing policy and service to his customer is
    really irrelevant to the question at hand, except insofar as its
    side-effects affect you.

Yup, it isn't my business to care about his internal routing, just to insure
that he presents appropriate routing to me (i.e. I care about the functional
specification not the design details...)

I was merely speculating, perhaps incorrectly, on what might be going wrong
so that I don't see what I consider consistant routes.

    Ah, I finally see the problem. In essence, Randy's (announcement) policy
    assumes that a destination is either in the set of customer routes OR the
    set of peer routes, but not both. But in this case we have a destination
    which is in both. The side-effect ends up making you "cold-potato" route
    to destinatons in the (customer + peer) intersection.

    The question is, does it make sense for a destination to be considered to
    be in both sets? If "peer" is supposed to mean "not-customer" then the
    answer is probably no, there is (should be) no intersection between
    "customer" and "not-customer".

See your objection above - I really don't care how the routes are handled
internally to Randy's net; I just want to see routes for the same prefixes
with equal preference (as path length, origin, etc.) at all interconnects.