Notice in my equation in the previous post, I also mentioned "Game direction cost". What I meant by that is the "cost" in terms of taking the game in a direction the developers do not intend and don't like.
For example, Talad doesn't really like "/tell" commands because he thinks they are unrealistic--and he wants this to be a very realistic game in certain ways because it makes it more immersive, which is what he is really after. The contrary position is that "/tell" is an easy feature to implement and incredibly useful to the community, social bonds, friendships and cybering, so it is "worth it" to sacrifice some realism or immersion to put in the feature.
Pickpocketing, OTOH, really is against our philosophy that one player cannot *involuntarily* or *anonymously* do damage to another player. Zanzibar's idea actually does get around this principle. By spawning a fresh item, the thief has the sensation of stealing while the "victim" isn't really hurt. For small enough items, this probably works. Inevitably though, if we did this, thieves would rapidly demand that their progression was "pointless" unless higher level thieves could steal bigger, more valuable items from players. The more we allow, the more valuable "something for nothing" becomes until the skill messes up the economy because it is all anyone does. Could we then nerf it by adding npc guards who constantly monitor and a system of fines, prison terms, etc.? Yes we could. <Insert reminder of coding effort in "Cost" equation here.>
These are just examples. Every feature ever considered basically has this sort of analysis done on it.