I agree, Mark. Remember we once contemplated "How long is too long"? What was the upshot of this? I haven't yet seen the rule, but I may have missed it - was some definition of an "upper limit" defined?
For instance, we sometimes find ourselves concerned about how a particular race organization may or may not do a satisfactory job of course layout. So, we assume that some of its participants may run the tangents despite the RD's assurance that he will place a line of cones down the middle of the road to keep runners on a single side as they run both directions on a stretch. So, we measure the full tangent. If the cones are laid out and runners stay on one side or the other, then they are running longer than the measured distance, but not shorter. So, we consider that we have done our job - right?
Now, consider a course such as G.B. mentions, with a long unpaved section, but now without defined edges or repeatable features. Most of us have encountered these, right? Could we, in theory, at least, measure such a section, assuming no nearby tall buildings and good sky view, using GPS coordinates to define the path and the timing points? We could, again in theory, it seems, calculate the largest likely error distance of each recorded GPS coordinate based on the particular average error distance of the instrument we use to obtain readings, add up all of these variances, and extend the course by this amount. The course as run might be long, but maybe not much more than in the example of a road course divided by a traffic cone line, right?
Yes, I am engaging in "devil's advocacy" with these questions. With a purpose. For instance, I once mentioned on this forum a particular event in which several national records have been set. A validation measurement revealed the course to be over 45 YARDS long. This creates an unsatisfactory situation in which another race of the same distance, validated to say, 12 inches long, could produce a faster record performance by a slower run, right? The "long" course in question is advertised and certified at the same distance as the one validated at 12 inches long. Now, 45 yards at a national-record pace is only a few seconds. But I argue that this record discrepancy is real, not theoretical.
My purpose here is twofold:
1. In the real-world case of the event referenced above, the course is not certified at the measured distance. It is certified at the nominal standard distance. A validation failure is a situation that none of us ever want to be in, obviously. My sense is that the measurer knew that this event had produced national records, even one world record, and he essentially "padded" his overall distance so that there would be no chance of a validation failure. If we measure a course to be longer than the advertised distance, and certify it for the measured distance, should we have some kind of upper limit past which we should not support the naming of an event at the (different) nominal standard distance? This is another version of the "how long is too long" question. Is this unimportant, or should we have some % standard?
2. I am on the side of at least some XC courses being certifiable. My reasons are partially explained by the arguments above. These arguments are supported by the fact that calibrating on paved certified cal courses, as we do, then measuring on an unpaved course, tends to produce slightly longer courses. If we are OK with a divided-road course potentially being long, we shouldn't be averse to an off-road course being equally long - right?
This discussion tends to lead me into a separate, if related topic - measuring off-road courses. Rather than invoke a reprimand by our esteemed Admin for conflating or discussing multiple topics in a single post, 
I will post some thoughts on a different thread.