It is desirable that if diacetyl is being made, that only your brewer's yeast culture is doing so. Most of the time, any diacetyl produced will be reduced to acetoin (by either enzymes that generally reduce any small ketone to an aldehyde, or enzymes specific to the task). This is the process labeled 1. In biochemistry, reduction is commonly done by using adding two H atoms via NADH + aqueous proton + 2 electrons.
A second reduction will form 2,3-butanediol (not shown, that reduction reaction is indicated by the unlabeled single down arrow below acetoin) The carbon with the -OH group is chiral. In 2,3-butanediol the second alcohol group is where the remaining ketone (C=O) group was, and that carbon also is chiral after the second -OH is made.
A large source of problems with diacetyl in the final product, is high levels of alpha-acetolactate which can convert to diacetyl in the oxidative decarboxylation process labeled 2. Generally, you can reduce formation of diacetyl in finished beer by decreasing process 4, increasing process 3, or increasing process 5. You can also lower the amount of diacetyl in the final product by making more diacetyl (should be a line in Alanis Morissette's song Ironic, don'tcha think?). In other words increase the rate at which reaction 2 occurs, as long as reaction 1 removes the diacetyl so produced. The latter approach (reaction 2 followed by reaction 1) is done intentionally when doing the diacetyl rest in lagers(raising the lager to room temperature for a couple of days after fermentation is complete). If you judge a lager with detectable diacetyl, absent any other off-flavors that indicate an infection, drop a hint that their diacetyl rest may not have produced the results desired. Don't say "You fergot you're diassetyl rest, dumbass!".
What is not obvious is that if the yeast is not provided with enough valine, the pathway which makes valine from pyruvate (top of figure straight down to bottom) will be upregulated, producing higher levels of alpha-acetolactate.
Of course, different strains of yeast will vary in the amount of diacetyl they tend to produce even under optimal fermentation conditions. Highly flocculent English ale yeasts tend to produce higher levels of diacetyl. This is why English Ale styles in the BJCP guidelines says that some diacetyl is acceptable.
A last note on removing diacetyl or preventing its formation: adding enzymes such as SEBmature L or Alphalase Advance 4000 or alpha acetolactate decarboxylase (an enzyme that accelerates process 3.) I think you can google that for yourselves if you got this far in the post.