During the early days of the V-22 program, I worked in the structural dynamics group at Boeing in PHL
. Most of our NASTRAN runs would run on a detailed model of the fuselage(what we built) with super simple "stick) models of the empennage, wings, stow assembly, nacelles, etc.
These would usually run on an IBM mainframe and in those days cost about $3-5k. These would regularly have mistakes and need re-running, bad but not terrible.
Periodically(every 3-4 months), we would merge our detailed model with those from Bell, Grumman, etc. and do some runs for this much larger, much more expensive model on the Cray in Seattle. We had a guy working on contract from Cessna(few jobs there in the mid 80s) who screwed up one of these runs with a bill over $100k. He was not a happy camper but we had a great boss and it was treated as an educational experience for all and I can assure you that he never made that mistake again.
Oh yeah, I've also seen a few gear up landings and a few people who had bad experiences with carb ice when they forgot heat on landing.
[Edited 2005-11-14 23:59:25]