I don't quite understand your objection here. Ordinarily the rule of the game is "when you buy a card, you gain it". What this card does is replace that rule while it's in play.
(Also, a clearer way to write your revision would be "when you would gain a Curse due to buying it...".)
"When you buy a card, you first buy it, then gain it" is the rule according to the Hinterlands rulebook. It doesn't replace the rule, it just replaces the gain. It would have to say "when you buy a curse, instead of following the rule 'when you buy a card, you first buy it, then gain it', each other player gains a curse" in order to replace the rule. And there is no gain being going on at the time you buy a card, so there is nothing to replace, and that means nothing happens. Your wording for my revision is admittedly better than mine.
If you have two "on would-gain, ___ instead" effects, does the first one you choose to resolve prevent the other one from resolving? Possession reacts to would-gain, so can the possessed player reveal a Trader to get a Silver (which the possessing player gains instead)?
Obviously yes, because you have an unlimited number of those effects whenever you would gain a card and have a Trader in hand, because nothing prevents you from revealing that same Trader 9001 times as a reaction to the same would-gain. With Possession, if you buy a Curse and reveal a Trader, you can choose the order and if you choose to apply Trader's replacement effect first, Possession's replacement effect fizzles because that would-gain no longer exists, but then Trader's ability makes the possessed player would-gain a silver, which makes both Trader and Possession go off again, and since there is no point in replacing that silver with a silver, the possessed player probably doesn't reveal the Trader this time, so there is only Possession's replacement effect left, and that makes the possessing player gain that silver instead of the possessed player. But if you choose to apply Possession's replacement effect first, Trader's replacement effect fizzles because that would-gain no longer exists.