EDIT 2: Creating an arbitrary distinction between "combo" and "combo deck" is even sillier than narrowing the definition of "combo." Do you mean to tell me that most combos cannot form a combo deck? That's even more confusing!
Dude, think in practical terms. People have a natural intuition of what a combo is, and trying to globally redefine the word is a nearly impossible task with no real benefit for success. If you want to use the term "combo" only to refer to a subset of card combos, you go right ahead. But telling everybody else they're using the word wrong is a losing battle. That's why the burden to come up with a new term for "combo that is an entire strategy" is on the people who desire such a term.
This particular line stuck out to me. You guys are mostly talking about how combos in Dominion relate to combos in MtG. Then someone brought up food deal combos, which doesn't evoke the same thoughts as MtG combos. The term "combo" is very prevalent in fighting games, and combos are more associated to fighting games than the way it's used in MtG and Dominion.
In fighting games, combos are basically sequences of moves that, when executed under certain conditions, cause something special to occur that would not occur if the moves were executed in isolation. For example, a move is automatically added to the move chain, or the last hit is always critical, or every move in the chain is guaranteed to connect if the first one does.
Hearthstone's "combo" mechanic lines up with this concept of the combo. Combo effects on cards are only activated if the card is played after another one, and never execute if the card with the combo effect is the only card played in a turn.
Now with Dominion, I think of combos as card interactions that result in strong assertions that cannot be made if the cards are played individually. These "strong" assertions are things that are more than just a card being better at what it's expected to do.
Non-Combo Example: "Play Plaza, discarding a treasure for a coin token, then play Library to draw up to 7."So here, the Library mitigates the drawback of having to discard a card with Plaza. But after playing both cards, I have a coin token and have 7 cards in hand. This is something I'd expect to have as a result from the union of the effects of Library and Plaza. Mitigating weaknesses in this way is a synergy.
Combo Example: "Play Rebuild, naming Tunnel, to reveal the entire deck without finding a VP card not named Tunnel. Every revealed Tunnel is discarded to gain a Gold."If Tunnel is the only VP card in my deck, then I can assert that playing Rebuild, naming Tunnel, makes me gain a Gold for every Tunnel in my deck. This is not just more than the sum of Rebuild and Tunnel, this is creating an event that should not result from using these cards under most circumstances. In fact, just having a single Estate in my deck makes the "gain 1 Gold per Tunnel" effect no longer hold.
Another Combo Example: "Play King's Court to play Scavenger 3 times. Discard the deck on the first play of Scavenger and put King's Court, Scavenger, and Silver on top of the deck. On the next turn, buy Province and repeat the King's Court->Scavenger play."By playing KC->Scavenger to topdeck KC, Scavenger, and Silver, I can assert that I'll be able to buy a Province every turn, so long as I have extra copies of those cards in my deck after cleanup and draw. As strong as it would be to topdeck 2 KCs and 1 Hunting Party, that doesn't guarantee that I'll be able to buy a Province next turn.
With NV/Bridge, if you reach the critical mass of Bridges on your NV mat, then you just empty the Provinces, most likely winning the game. Same with Hermit/MS if you reach the precise balance of Madmen/Hermit/MS. That stuff doesn't happen if you replace NV/Bridge with Highway/Worker's Village and Hermit/MS with Hermit/Market/Ferry (making Market cost 3 to gain with Hermit). In short, these combos create events, significant events, outside of what is reasonably expected by simply resolving all the effects on these cards in some random order.
Tactician + Black Market is kind of in the middle. I lean towards considering it to be just a really, REALLY good synergy. The result you get from playing both is what you'd expect from the cards being played individually and then those results are mashed together. However it's still reasonable consider "play all treasures for value despite playing Tactician" to be a significant event outside of what Tactician and BM normally provide, thus making it a combo. Tactician/Vault however is definitely a combo if you can guarantee to draw Vault and Tactician every turn, as this leads to a Province every turn.
Following this idea of combos, Golden decks like those possible with Bishop also fit into the definition of a combo, analogous to endless chains in a fighting game (presumably broken if an endless combo chain is possible).
If every synergy was considered a combo, then engines would just be a long combo chain. Some players are bound to actually see engines this way, calling engines "combo decks". I think people I've played with have described them as such. For us on this forum, we see combos as interactions that lead to the same results you might expect from an engine with what seems to be way less.
So ultimately, I do think the term "combo" should be restricted to an interaction producing a result at a certain power level. Not only that, that result should be something that feels almost like you entered a cheat code to get something that you wouldn't expect at all from standard use of the cards that are part of the combo. Notice that a lot of combos are dependent on rather restrictive conditions being met, like having the right amount of Madmen compared to Hermits for Hermit/MS and having Tunnel be the only VP card in your deck for Rebuild/Tunnel. From there, you can classify between viable power combos (like NV/Bridge) and support combos (like Rebuild/Tunnel).