Wait, a 2% chance is fairly high (I would think) for a spacecraft. How often would this possibility of malfunction be computed?
For an XR-class vessel, I agree. A futuristic vessel like these I think would have the same reliability as modern airliners. According to this presentation,
http://ocw.mit.edu/NR/rdonlyres/Aer...8E3F-4339-B67E-664318D46F41/0/reliability.pdf that seems to be about 0.01% per flight. If we convert it to reliability per hour of flight and call it an average 3 hours per flight, that .003%/hr (~8E-7%/sec - how often I would compute it). Lets further decide that any system failure results in an 'accident'. The Earth moon run is about 7 days (lets round up to 200 hrs). We're still above 99% reliability. Mars at 3 years (26,280 hrs), on the other hand, has about a 70% chance of failure. This isn't quite accurate, though, as most accidents occur at takeoff and landing, of which there is still only one of each per trip. One percent seems high to me per trip, as the Space Shuttle is at about 2%.
Another idea would be to put the failure chance in the config, so the end user can change it (sort of like the XR series' "cheat codes"), so if the end user thinks that it breaks down to much, they can change the percentage/probablilty.
This is the best way, as people wanting to do station-building, for instance, could set the level at a more 'realistic' level, and those that wanted to practice emergency procedures could set it high. It would also be a good idea to set reliability values for each system. Avionics would be very reliable, but mechanicals/hydraulics or TPS would be less so.
The key is that if failures are possible,
there has to be something the pilot can do about it. Otherwise, why simulate them?