Mike:
The attack posted here uses a brute-force search to find a phony LEAF which has a valid checksum. Instead, why not just initialize the chip with a session key and get the LEAF. Reset the chip and initialize it with a different session key, but send the first LEAF instead of the second one. The LEAF would look good unless you tried to decrypt the session key. The wrong-IV problem would remain. The NSA should have designed the Clipper so that, if the IV was wrong, the chips would not accept the LEAF. They also should have used a much larger (32-bit or even 64-bit) checksum.
Because if *your* key really generates the LEAF, then they have your ID in the LEAF, no matter if it is sent properly or not. They might not be able to decrypt the communications, but they still get your ID. If you randomly generate a LEAF that works, odds are that the decrypted value will not be your ID. (If you could consistently choose random blocks such that your ID appears when it is decrypted, I would say that you have found a hole in Skipjack :-) -derek