Elizabeth Anderson has a post up at Left2Right about distributive justice, in which she argues that distributive justice should be seen as procedural, because patterned or outcome-focused justice requires excessive intervention on the part of whatever body administers it, or, to put it another way, the infamous claim of Nozick that 'liberty upsets patterns'. In particular, she argues, contra Nozick, that Rawls's theory of justice is consistent with the Hayekian injunction against patterned theories of justice. I think at least has a grain of truth in it, because, while there might be an argument about the difference principle, the other one and a half principles of justice are clearly procedural - the widest degree of liberty, compatible with equal liberty for all, and fair equality of opportunity do not mandate any outcomes at all, but merely tell us how any just outcomes will be produced - and the original position is one of the clearest expositions of the idea of procedural justice going.
Rawls himself draws a distinction between three types of procedure: pure procedures, procedures which require no idea of a just outcome in order to justify their outcome - tossing a coin or a lottery, for example; perfect procedures, procedures which always achieve an outcome which, independent of the procedure itself, we believe is just - the practice of having one person cut the cake, and the other choose the slice, for example; and imperfect procedures, procedures which aim at but do not always achieve an outcome which, independent of the procedure itself, we believe is just - criminal trials, for example. There may also be impure procedures - most team games would appear to be of this sort, because there is no independent standard of who ought to win, but we can intelligibly say that 'x didn't deserve to lose' - although Rawls doesn't mention them.
I'm a little skeptical about the existence of any genuinely pure procedures - one might describe a lottery as aiming at a distribution in which each ticket had an equal chance of winning, which would then make it an example of a perfect procedure, and I think that kind of redescription can be done for any putative pure procedure - which would mean that, whilst it may be true that 'liberty upsets patterns', no theory of justice can get by without some notion of how things ought to turn out. This is not to denigrate procedural notions of justice: personally, I think that the description of criminal trials as imperfect procedures is a little off the mark, because it seems to me that part of the outcome they aim at is specified in their procedure. It is not only that the guilty person be convicted, but that they be convicted by means of public evidence in a public forum, where public is clearly procedurally oriented: unless someone can be seen to be convicted, they should not be convicted, which is a matter of procedure, not outcome.
What both of these examples do is cast doubt on the idea that the requirements of justice can be easily articulated in a language either solely of outcomes or of procedures, or even in a mix of both those languages: they suggest that the distinction is not as sharp as might be thought. I'm not sure what the implications of this are for Anderson's discussion, because she doesn't make it clear what kind of procedure she is talking about when she describes justice as procedural, but it does seem to cast doubt on the libertarian idea that the market is a pure procedure, and justified because of that (that's not to say there aren't other possible libertarian arguments, of course).