Granite has a good point. Some of us want to look at the walls we built to contain magma and think to ourselves, "Are these strong enough? Maybe I should make a second layer of walls. Out of ice or something." Others want to say "These are walls. They won't break."
So the failure rates and effects (output) should be in the raws, so that people can make a... er... "lego" mod.
I'd think that this should work for everyone--hammering out reasonable failure effects should be the next step?
No. Wrong. Bad.
Everyone reading this knows that there will be one 'vanilla' dwarf fortress around which the rest is built. New things will be balanced for it; bug reports will only be at their maximum usefulness if they come from it; new features may be introduced or tweaked based on what it already contains. You cannot "offer" other people the option of having the existing relatively smooth, comparatively micromanagement-light, failure-free gameplay confined to a mod as an olive branch; if you want to resolve the dispute, suggest that your
own desired tweaks be an optional mod on the vanilla, fail-free Dwarf Fortress. The distinction is nontrivial, especially when dealing with a suggestion that so many people clearly feel is so wrong-headed; this is not a suggestion that should contribute to the 'core' development of vanilla dwarf fortress in any way, and is not something against which anything should be balanced in the main game.
For the reasons countless people have explained here over and over again, random failure would be a pointless, unconstructive annoyance that adds nothing useful to the game. It would not pose a serious challenge that requires strategy or planning or anything
interesting to overcome; indeed, all it would to is enforce dull, dross micromanagement of tasks and workshops, adding pointless and elaborate steps of micromanagement every time you want to train a new dwarf, and forcing you to carefully segregate all labor to avoid catastrophic failures.
Even the people who love the degree of pointless micromanagement that this would produce admit its core, irreparable failing in trying to defend it -- they say that it would have no effect on the game as long as you constantly remember to appropriately manage every new workshop (and, presumably, micromanage where every new task you assign ends up, and never use the manager screen, and constantly remember where your real and 'training' workshops are for your most trivial tasks, and dozens of other annoyances...) They want all this for something that they themselves admit would add
nothing to the game once you have skilled dwarves for important tasks (hint: you can start with skilled dwarves for important tasks.)
This admission deals an absolute death-stroke to their own bad ideas. If these horrid random failures of yours can be easily managed, what purpose do they serve in the game? You are requesting a substantial change -- one that, at the very least, requires more mindless annoyances for the player to bother with at some point -- while simultaneously trying to make the argument that it will have no impact on the game. You essentially admit that this offers no serious possibilities for creativity nor an interesting challenge for those interested in a game; it is just a dull step to be followed by rote when setting up workshops, which can then (by your arguments) be totally ignored.
An idea as badly-thought-out as that obviously cannot become the default assumption of the dwarf fortress universe. If you want to request that there be an
option for masochists and micromanagers who wish to mod vanilla, failure-free Dwarf Fortress to have random failures, go ahead and do it; but offering to turn the sane and fully-functional game that we have now into a mere mod of your "dwarf-butterfingers" default is no compromise at all.
Or, for the TL;DR version: Your proposal is fine as long as the default 'reasonable failure effects' we hammer out are 'NONE', and the probability is 'ZERO'. People who view additional micromanagement as a challenge and who wish to address their creative abilities at mechanically assigning workshop minimums can then raise them on their own personal copies of the game to whatever they desire.