Make a case for keeping it.
How about "most of the skills are under-used, as the mod is *really* combat orientated, so eliminating the benefits conferred by a skill that IS used only lowers variety on the server more." ? Not a great case, but a rather solid one.
My n00bish opinion on the matter:
PP *is* PvP. It is a hostile action taken by one player against another. The only dispute is a question of scale. As the value of the object stolen could theoretically be higher than the value of the gold lost from a PvP kill, the case could be made that, in terms of "injury" caused to a player, it is substantially worse. Similarly there can be instances when the XP lost from a PvP is negligible compared to the value (subjective rather than monetary in this example) of the item.
Saying that "PP can be used to legitimately provoke players" is a total farce, has nothing to do with any sort of "justice" (frontier or otherwise).
Insulting someone's mother can be used to "provoke players" and its "legitimacy" as a method is only *arbitrarily* different.
What people are basically saying is "We want a get out of jail free card to start otherwise illegal fights and pick-pocketing, and its place in rules-limbo, gives us an inscrutable vehicle to do it."
As Ironfang rightly points out, if you think players have the right to "sort it out themselves" - you may as well give them the right to do it with PvP instigation, rather than them instigating it in the round-a-bout proxy way of doing it via pickpocketing someone.
As a noob, I've not noticed a problem with pick-pocketing, although someone in my party implied I was doing it to them, despite being in the same party in the middle of town (non-PvP zone, right?) - an area I assumed it is impossible in?
*Ideally* I'd like to see protection from pick-pocketing in the form of some feats being taken into account when rolling an opposed spot check.
IE: Work out the max amount a viable (non-dedicated PP) build could have in it, and call that 100% PP. Work out the max amount a dedicated spotter build could have, and call that 100% detect.
Code it so a PP can have a couple of results:
Pick-pocket is spotted and fails (nothing is stolen)
PP is unspotted but fails
PP is Spotted but succeeds
PP is unspotted and succeeds
Use the equasion
If (PP + 1d100) - (Detect + 1d100)
is <50 there is a critical failure (similar to critical failures in failing to disarm traps and accidentaly setting them off, or antagonising animals with failed empathy, etc) and the PP is spotted and fails.
If it is >50 but <100 he fails, but isn't spotted.
>100 but <150 spotted, but suceeds,
and >150 he is unspotted and suceeds.
The beauty of this system (the numbers I gave are arbitrary just to illustrate how it could work) is that the variables can easily be adjusted to create different effects. I distributed the 4 effects evenly just because it is nice and simple to do so. Likewise, I gave the percentage and the roll equal weighting just for clarities sake, and because it means that at low levels it is much more random, reflecting the ineptitude of thieves at that level.
To make thieving more difficult, it is quite simple to lower the PP 1D100 to say 1D75, meaning a thief would have to have substantially more skill in thieving to be succesful. This can be used to balance out a number of things, such as eliminating the massive bonus Rogues get from having PP as a cross-class skill and lower the success rate to 1/3 (all things being equal, etc).
ANOTHER thing you could do is use some key feats to give an extra invisible bonus to detects. For example use the awareness feat, listen skill, luck of heroes, etc etc to add to the detect die roll by a set amount. *OR* take into account saving throws SOLELY on the side of the person trying to resist the theft, giving them a reflex roll bonus to the 1d100 roll.
If you wanted to get REALLY sophisticated, you could balance it on the other side by making the silver-palm feat give a similar bonus to the PP die roll or even add to the equasion so that there is a roll for succeed fail along the above line, and a seperate roll for detect / not detect based on stealth checks, stealth feat, etc.
All these things would detract from the "min-maxing for combat" cookie-cutter builds that I seem to keep running into...
It can be as sophisticated or as simple as you are able to code it, and still be "fair and balanced".
Of course, I have no idea how feasible it is to recode how skills work against each other, etc.