I think I'm getting comfortable enough in the 747 to add some spice to my flights, i.e. the random simulated failures. Apologies if this was answered in the intro manual, and if it is, I'll head back there, but this is my go-to place for asking "obscure" questions. Feel free to point me to the docs, if these questions are answered there. 

In the FMC, I've enabled random failures for "All Systems" that should be happening at a rate of 2x per 10 hours, limited to a single failure. I also have service based failures enabled. This I have enabled a few days ago, and I had one engine flameout (nice!), and aileron lockout (didn't even notice, until after I looked in the FMC post-flight if something failed). So, at least some of this seems to be doing something. Now for my actual questions.

1. Browsing through the failure options in FMC, I noticed that while "All Systems" has failures enabled, as described above, individual systems seem to have random failures disabled. I am *guessing* that "All Systems" overrides individual systems in this case. Is this correct? OR - do I need to enable individual systems for failures, before my global settings kick in (based on my so-far experience, the answer should be: no).
2. Are those X failures in Y hours, per "sim time", or actual time? I.e. if I'm using intelligent acceleration at 8x, I should be expecting a failure about every ~2 hours or so. This, I have not observed.
3. Are those X failures in Y hours per cumulative flight time, or counted for each flight?

 

Last, but not the least: any chance a failure could trigger revert to normal time during the accelerated flight? That would be really nice, as it took me some time to notice that flameout :-)

Check page 48 and page 116 in the intro manual for service based failures. This should answer at least your last two questions.

Thanks, Kevin. I've seen those, but as I said in the original post, I am still a little unclear :-).

I can help with parts of the question.  Each aircraft by tail number keeps track of run time, and if you look into that configuration file you'll see a bunch of run timers. The time accrued is simulation time so time compression will build time faster.  I'm not sure how the failure rate is modeled but if it is a realistic Gaussian distribution and your failure rate is once per 10 hours then it is statistically reasonable to not get a failure for more than 10 hours then suddenly have a rash of failures.