GW1 spun out of control in a way, because of power creep and because the number of combinations were too many to balance. The MtG method is in a way self-balancing, and that is a very important concept to horizontal design. Another example of self-balancing mechanics would be some of the most well balanced skills in GW1:
Frenzy: Probably the best general IAS in the game, but one that viciously punished you it if you use it at the wrong moment. Not only does this self-balance against the user, but also against the opponents, who are (doubly) rewarded for keeping a lookout for people using Frenzy.
Reversal of Fortune: Like any prot, this one is wasted if cast on a character that doesn't come under attack. But even more so in this case, because the skill has a potentially massive heal that triggers only if you cast it well.
Diversion: The skill that pretty much defines much of what mesmers were in GW1, it yields a massive reward if used right and is just a waste of its quite long casting time if used wrong.
In all of these cases, the actual numbers were not what kept these skills balanced, it was the mechanics behind them that did it. Of course, a 100% IAS in Frenzy instead of 33% would be unbalanced, but removing the double damage would be a much worse hit against the skill.
Why am I blabbing on about skill balancing when the question was how much horizontal progression there should be? Aha, because it's intimately connected. There must be self-balancing mechanics involved. There must be tradeoffs. Tradeoffs that feed back into the use of the skill, the construction of the deck, etc. If you run out of tradeoffs, or if they don't work, you should stop putting out mechanics. In other words, there can be as much horizontal progression as the mechanics allow, and there can be as much mechanics as the tradeoffs allow.
Very good post. And what you describe in your last paragraph is one of the main Problems of horizontal Progression: Every new Option requires meticulous balancing, something that takes a lot of time and often requires innovative thinking. There is a sharp limit on how much horizontal content you can make, and that directly conflicts with the design philosophy of current MMORPGs, which revolves around providing as much content as possible to keep people playing.
Maybe a major design shift away from the "Theme-Park-MMO" towards a user-created-content mentality could help (read about that recently).
Nevertheless, I think a pure horizontal focus will not work in games with a PVE focus. The appeal of PVE is often based around being able to beat tougher foes, the method is not terribly important. Horizontal Progression works in player-player Interaction, be it PVP or showing off skins or titles. MOBA's have purely horizontal progression systems, and if other genre's focus more on player interaction instead of PVE, we might see more of it.