It's not so much a question of "do we have failures or don't we", but rather how deep the system simulation is. To me, deep system simulation means the systems need to behave realistically when things go wrong or when a pilot mistake is made. Like others have said above, what's the point of coding deep system fidelity if nothing is ever out of the ordinary. And by that I don't just mean catastrophic events that force you to land immediately. It's more like the little things that happen on a daily basis in airline operations. Chances for a hot start, HYD pressure that can drop for whatever reason, INS issues, a pack that runs hot... just things so that system monitoring and checklists have meaning. Without that, it's hard to enjoy deep system simulation, because there's no reason for it to exist when no parameter can ever go out of line.
So yes, please, add system failures, errors, acting up, randomness. I have no problem with paying extra for this. It's just that flying with knowing nothing can ever go wrong is, how should I say, not very fulfilling.