I don't play a game with alignment. The notion of "NG" does no work at my table. The PC in question is neither NG nor not-NG, because those descriptions are not part of the framework of my game.
Whether we slap the NG label on the character or simply allow the player to say his character is a decent fellow who does the best that a good person can do, is devoted to helping others and works with legitimate authority but does not feel beholden to them - he believes in doing what is good without bias for or against order, we still have a description which we cannot gainsay. If the player decides his basically decent character decides to take any given action, we must accept that the action is benevolent and decent, since the PC's moral code calls for such and he can never be outside that moral code without player consent.
If the paladin chooses to destroy it, then it was not a treasure parcel for that player, and hence s/he can expect to find something else. If the paladin chooses to keep it, then it was a treasure parcel for that player. Why would the player of a paladin, who thought using poison was dishonourable, choose to cash in his/her treasure parcels in this way?
So if my character wants a very specific item of treasure, can he simply toss away whatever else he finds until the GM gives him what he wants?
In my campaign, there is a morally-laden cosmological question at issue - heavenly order vs primordial chaos. I don't think the alignment mechanics have anything to say to it at all, because the alignment mechanics do not tell me whether law or chaos is more desirable (and I do not see why a LN character should oppose a CN one any more than a peanut-butter eater should oppose a chocolate-eater). That's just one way in which the alignment mechanics do not contribute to my game.
I don't quite grasp how we have an order versus chaos conflict, if the lawfuls and the chaotics get along fine. Conflict implies they don't get along so well.
I reiterate that you have no evidential basis on which to judge that alignment is not an impediment to my enjoyment of those episodes of play. I aslo reiterate that you tend to conceive of "an episode of play" purely in terms of the fictional events that occurred during that episode, whereas I am referring also to the emotional and other experience of the participants. The same fictional events could have been achieved via GM railroad, but that doesn't mean that the presence or absence of GM railroading has no bearing on my enjoyment of the play experience.
Absent the same play with and without alignment, and a comparative study, we have no evidential basis for any specific element adding to, or detracting from, play.
I don't know - I'm not the one who uses alignment.
You seem to frequently switch between having no idea how alignment works, as you don't use it, to being expert in its usage to determine how others would apply it and how that application would detract from your enjoyment of the game. Which is it?
\PCs in 4e are not immune to being persuaded. But the mechanical resolution is different - the player gets to decide (whereas for NPCs the matter is determined via skill checks).
Sorry, but where the player decides, the PC is de facto immune.
As to loss of hp being comparable, we've danced that dance more than enough, I think. Do the hit points just vanish with no mechanics behind their removal, or are they based on actual rolls failed by the player, or succeeded at by his enemies. You have consistently provided rules quotes on the consequences of failure to support the adjudication being simply part of the rules, while steadfastly refusing to indicate what roll the player failed to result in a negative consequences you imposed. As such, my opinion remains unchanged.
But, you the DM, wearing the hat of "cosmological forces" has to make a decision. How can it be good and not good at the same time? That's exactly the problem you accuse me of by not using mechanical alignment in the first place. It's inconsistent.
It can be both "not good" and "not evil", just as it can be neither chaotic nor lawful. It would be inconsistent only if the same action, under the same circumstances, is good sometimes, evil others and neither at other times. It is also inconsistent if the action is categorically Evil for PC 1 and definitively Good for PC 2. To judge that circumstances have no impact on judgement of the action also invites ridiculous results. Is it Good to kill people? In a vacuum, I must say no. But the decision to kill is not made in a vacuum, and the context must impact on the judgment of the action. You seem to want a simplistic check the box "this is good, that is evil" model, and I reject that as being a reasonable implementation of alignment in any form