Right, I agree that a card that varies too wildly with each kingdom would be bad. But I'm not convinced that this "cursed gold" (+$3, -2VP) varies that much.
If there's trashing in the kingdom, such a card is basically a Gold for $5. You buy a bunch, then trash as many as you can before the end of the game, and they might as well have been actual Golds. Considering other $5 Gold equivalents have steep penalties, this is clearly a power card. So maybe you crank up the VP penalty, to introduce a bit more risk. Now look at games without trashing at all. Would you really ever buy a Gold that put you down by a Duchy with no way to recover? Maybe in Village+Goons games, where the point counts get so huge the penalty is dwarfed by comparison. But in a normal Province game without trashing, probably even -2 VP is too steep.
To be sure, there will be boards where the balance is right. But I'm pretty convinced that this is a very small percentage of boards. And on top of that, there is also wild variance within games based on when you buy the card. Buy it early, and you use the Gold a lot; buy it late, and you don't, but you still get the same penalty. Sure, other cards are timing-sensitive too, like Chapel and Trading Post. But those don't vary wildly on TWO axes, just the one.
Because the problem is that the benefit and the penalty don't scale with each other. In a long Goons slog, you see the Gold a lot, but the -2 VP penalty only happens once. There is no default point of balance. It's either too much of a penalty, or too little, and only just right by accident.
Contrast the flat one-time -2 VP penalty with Cache's two Copper gain, or even the penalty of gaining a Curse separately. Deck clog is a penalty that scales with the benefit. The longer the game goes and the more you shuffle, the more you see the Gold AND the Coppers/Curse. Buy it early, buy it late, buy it on a fast board, or buy it on a slow board -- the benefit and penalty will have a tendency to balance each other out, because the two sides aren't fluctuating independently of each other but in sync. There is still PLENTY of variance, because different kingdoms will have different means by which to cope with the penalties. Presence of trashing, sifting, etc, will help, while single-card chains will choke. Envoy will penalize high-variance decks, while Oasis rewards them. Stuff like that. But the balance is anchored around a sensible average.
The fact that the Curses easily run out, though, is reason enough to avoid the "When you gain this, gain a Curse" variant. I'm almost positive that was a consideration when Cache was designed.