It's in the first paragraph of WW's article.
A combo deck is a one that revolves entirely around a particular specific combo of 2+ different cards, generally getting 5-20 copies of the required cards in total. Once the combo is in place, if this has happened quickly enough, the deck should basically just win.
Yep, that's the part where he kind of hints towards it very slightly. It doesn't say anything about the early game consisting of setting the combo up, which is the main part, and that's also probably why he mentions Apothecary/Native Village even though that's an engine, not a combo, as it spends the early game building an engine.
Even with Zakharov's definition, it's still about setting up a combination of cards. So are you recanting on your earlier definition that includes a "strategy" of nothing but the starting 10 cards?
It's worth noting that while both Zakharov and WW mention a combination of cards, they both also include the traditional Bishop golden deck in their list of combo decks even though that's a single-card combo.
I am recanting my earlier definition in favor of this one (and to be explicit, "this one" = "a strategy that involves spending the early game setting things up, until things are set up") because I like this one more and they are functionally equivalent for practical applications, but not the fact that a strategy that involves buying nothing is presumably a combo. I don't know why Seprix thought that it would be a good idea to go for such a strategy, but let's imagine, for the sake of argument, that it's because he knows the following event will get released in an upcoming expansion co-designed with ErrinF (which is why it's such a well-designed event and all that good stuff):
The Gunpowder Plot$2 Event
If it's your 11th turn and you haven't bought any cards this game, +100 VP and trash the Province pile from the Supply.
Going for a strategy involving this event is very clearly a combo.