I understand the Skipjack review committee will be looking into the key generation process at Mykotronx also. The procedures originally proposed for burning in the keys has some annoying flaws that have been pointed out frequently, like the existence of both halves in the same room at the same time, which would be a tempting target for somebody siphoning them off to a private single-site escrow. :) Various people have suggested that the two halves of the key could be burned in at separate locations, so that the only place they're put together is in the key itself; this was not part of the proposal as we've seen it so far out here.
..but what of at least the original description which had the key generated by a function within the sacrificial laptop (a PRNG driven by the two key pieces supplied by the escrow agents)? K_i = PRNG( E_1, E_2, i ) ; If PRNG is secret (a high power NSA algorithm), it might be a *truly* high quality one-way-function -- which happens to be independent of E_1 and E_2. No one would be the wiser -- and the NSA's job would be a lot easier, in the event that some chips get stolen or someone bombs the escrow center and its backup sites. - Carl