I see a strange distribution of the number of items of varying quality when making multiple items: At a high skill (97) I see most of the items being of top quality (as expected), usually a few fine (expected as well), never, or almost never, any decent quality item (odd), and then often some inferior quality ones (reasonable).
A typical example is a modded reaction to create bark cords where 14 cords were produced distributed as 12 perfect, 1 fine, 0 decent, and 1 inferior. I believe I've seen the same oddity with board splitting as well.
I think this distribution issue appears with single items as well, but it's not as apparent unless you keep close track of how many you're producing of each quality.
On the suggestion side, there are very sharp boundaries for quality. At a woodcrafting skill of 89 using fine tools you tend to get 0-5 perfect boards out of a split log, but at 90 suddenly most of them turn out perfect. There is at least one similar sharp threshold at lower skill levels going from an occasional fine result to most of them being fine at a single percentage increase, and I think there is one for occasionally->mostly decent as well.
This is probably related to the dismissed bug report of the first tanning hide working on a superior hide (with a sufficient skill) almost always results in either superior or decent hides, but very rarely in fine ones (after looking closely at it for quite some time I have seen at least one, very rare, case of just a single drop in hide quality though, so I agree it's not bugged as such, just a strange distribution of results).
What I would expect to see from quality distribution is a curve with a peak at the "expected" quality (which can lie in between the actual discreet quality levels) and then drop off on either side (which does not have to be symmetric). The "expected" quality level would creep upwards as the skill increases (but the process can be linear, logarithmic, or exponential, or something in between those), and the shape of the flanks does not have to be the same (for instance, as the skill increases, the lower than expected side of the curve might grow steeper, resulting in more "mild failures" and fewer "severe failures". One property of this kind of distribution would be that the number of items of a quality further from the peak would never be higher than the number of one in between that one and the peak (statistically speaking: weighted random results can of course give odd individual outcomes).
Indeed, it seems like decent quality result possibility was completely missing from above 90% mastery crafting. (Seemed almost like a typo on code level.) Fixed now, decent things for masters as well are on their way.
Fixed - persists in 3.40