You make a good point, but I'd go a different direction.
It seems to me obvious that terroir (and here I'm using my preferred definition) is an agricultural phenomenon. It's the chemical and physical makeup of the agricultural product that results from a given site's characteristics, differentiated from another site's different characteristics by the chemical and physical makeup of another agricultural product. Assuming identical farming techniques, plant identity, and plant age, what's left, as the agricultural product leaves its food source (tree, root, vine, whatever), is what terroir brought to it.
If that's the case, then it follows that an agricultural product must express terroir. It has no choice. Even if we abandon the theoretical crutch of controlled variables (farming techniques, etc.), the chemistry and form of the agricultural product is still influenced by the mesoclimate, the soil chemistry, and all the other inputs we include in terroir. It doesn't matter if pinot noir, syrah, chasselas, or turnips are planted on Romane-Conti...they will have, as part of their chemical and physical makeup, differences from the same thing grown on the Sonnenuhr. I would be surprised if anyone actually disagrees with this, in fact...but if they do, they're free to plant bananas in the Mosel or riesling in the Amazon and share their results, which might possibly show differences -- physical and chemical -- with the reverse case.
Because of this, I believe that grapes must express terroir. They have to. It's an agricultural certainty. There are (far too) many ways to mask that terroir, both in the vineyard and in the cellar, and obviously some grapes are far more transparent to site differentiation than others, but the grapes left the vine with a terroir signature of some sort.
I'm insistent on this as terroir, rather than one's ability to detect terroir in a finished product, because of all those natural and unnatural ways in which terroir might be obscured for the consumer of that product. If you took twelve identifiably different sites and planted cabernet sauvignon on all of them, controlling for every variable that can be controlled, a lab could show the difference where a palate might not, were the differences not large enough for the blunter and fallible instrument of the palate to detect. But the sites are still forcing differences in the grapes. And thus, there's still terroir. Some consider this a meaningless thing to assert because they don't believe it exists (or is meaningful) if they can't detect it, but consider this: uproot that cabernet sauvignon and plant sylvaner. Let's now assume the differences are apparent to both lab and palate. Did the influences of the site on the product just magically appear between plantings?* Of course not. And frankly, I don't want to rely on the fallible and variable instrument of the palate -- yours, mine, anyone's (well, maybe Claude's) -- to be the final arbiter of whether or not there's terroir.
* In Eric's conception, I think it did. Which is the fundamental difference between the way he and I are thinking about it: for him, there's no terroir without a product expressive of that terroir. For me, there's terroir whether or not the product expresses it. As I noted above, I prefer my definition not because I think his is invalid (again, how can it be?), but because it allows me to talk about what the other influences on a product (farmer, winemaker, clone, seasonal weather, ladybug infestations, etc.) did to the wine, without quite so much ambiguity about who did what. It allows me to challenge the assertion that a 16.2% pinot noir is the natural expression of a given terroir, because the other guy buying grapes from the exact same vineyard is taking fruit at 14.3%, and while both are making quality wine I obviously prefer one over the other. If I have no clear idea of what choices are forced by the site and what choices are actual choices, I don't have any grounds (other than reactionary ones) to challenge that assertion, and it's especially hard to draw other than caveman-level conclusions about the influences of land and hand.
That someone, anyone, or everyone might make a mockery of a terroir by the time the wine is bottled is irrelevant to my definition. (Not irrelevant to me as a consumer, of course.) The terroir is. As for the wine, that's a much more complicated discussion. But somewhere in there, whether naked and in your face or buried by MegaPurple, enzymatic nursmaidening, and crafted yeast, is terroir...its existence not invalidated because of what happened later, or because of my or anyone's inability to taste it.
As for that particular usage of "natural," I admit that it has always grated, because it's coming very close to redefining the word to mean "wines that I like." And if someone wants to use the word that way, OK for them, but then they're conceding that the word doesn't mean anything, can't mean anything, and hasn't ever meant anything apart from "wines I like," in which case it's of no use as a shared concept, and I wonder why we bother talking about it at all...whether or not some are recanting their usage. We already have an obnoxious way to refer to wines we like while pretending we mean something else: unspoofed. Why do we need another?
I take Eric's point on this issue to be a little more nuanced: "natural" or not, there's no way a grape pushed to and beyond its limits (whether by nature or by man) is going to express terroir as well as a grape to which this hasn't occurred, whether or not the winemaker is making what some might call "natural" wine. By his definition, this obviously follows, because he's considering a whole, finished, and cultural product. But even from mine, it does in the sense that we're talking about the expression of terroir (which I understand to be the ability to sense it in the wine), not terroir itself. I don't think the views are incompatible, in the sense that whether one embraces or rejects the tenets of "natural" winemaking, one is not released from the obligation to arrange the marriage of site and variety/ies with care, should one wish to make a good, terroir-revelatory wine.