Except that the section that Paul quotes is in Q.931 and NOT in H.225.0. It has nothing to do with H.225.0. In fact, it talks about a message that is not even supported in H.225.0 (i.e., RELEASE).
To reiterate: H.323v2 does not describe how an endpoint receiving both elements shall behave themselves. It just says that the sender shall not include both since it would cancel fast start. This particular case (wich apparently is not supported by any implementation) would result (if you follow the logic of H.323v2) in ignoring fast start and proceeding with H.245 tunnelling. That is NOT a backward compatibility problem, and it does't break anything.
Rejecting the call with cause value 100 is definitively NOT what H.323v2 says.
In Q.931, the user-user IE is typically optional, so the clause that Paul Long's been quoting wouldn't apply. However, since H.323 makes the UUIE mandatory, it *does* apply in the case under discussion. I think that the statement contained in TD-26, that it preserves backwards compatibility, is completely wrong. The correct statement should have been that it doesn't break implementations known to the authors. I agree with Paul that we should try to develop a better solution.