Well, I was thinking of it involving minimal human effort. Just trying the top 20 words in similar enigma messages maybe.
Maybe different lengths wouldn't work, but I don't see why trying cribs would hurt. I don't think it would take very long to try the top cribs.
Just a thought.

Trying different message lenghts helps a lot. Sometimes adding just few more letters (or removing some) changes a hard to break message into easy or at least 'medium' one.
I've done a lot of test runs on M3 messages last year using different algorithms and dictionaries, sometimes even padding the message with random text helped to break it, probably because it keeps the algorithm(s) from getting stuck at random positions with higher than average scores.
IMO the fastest way to break a message is to run ~30 walks through the keyspace, if it doesn't break, then save all the results and current state, change something and try again with different approach.
However it's not that easy when breaking M4 messages, because doing 30 walks even with 72 letters ciphertext takes a lot of CPU time. Workunit runtime increases with ciphertext lenght, for example a 150 letters workunit runs for about 2.5-2.7 times longer than 70 letters one.

Perhaps trying a biscore+triscore run on even shorter ciphertext (like 40 letters) would be a good idea. Icscore doesn't work very well on really short texts, but biscore with a good dictionary works just fine as first pass. Who knows, maybe with a good enough dictionaries it would be possible to break offizier messages, if the header is long enough.

Trying the cribs surely won't hurt and it doesn't require a lot of work, I've played a bit with bombe emulator some time ago, but without success.

---

M4 Project homepage
M4 Project wiki