So, I was thinking about the original point of the thread, which could be stated as the following hypothetical:
- Do we want one alliance to be able to dominate the game?
My impression is that we all agree, perhaps even wholeheartedly: we don't want to see that. 
So, I'd say we at least agree on the problem. Now, as to solutions, I can see three potential categories (with a fourth more exotic possibility; more on that later):
Caps and costs
Hard caps or costs can easily limit alliance size, and if alliances and guilds can only have limited members and towns, then it should be effectively impossible for a single alliance to dominate.
Personally, however, I'm leaning toward something Gothix has expressed elsewhere, which I'll state as this: rely as little as possible on hard-coded restrictions. Use them as a last resort.
Logistics
This is all about making it difficult to project power, difficult to show up with a thousand players to fight a battle with a hundred, difficult to steamroll an enemy city. Continent size, movement speed, supplies, fortifications, and so forth all have their place here, and a huge alliance would thus be less likely to dominate.
This is my favorite class of solution, and I haven't seen much disagreement (if any) here.
Tactical mechanics
This is all about making those battle odds of a thousand against a hundred much less uneven, so that the value of having a huge alliance is lessened. I would class friendly fire and collisions here, but it also includes things like ganking protection and attack spam resistance, and possibly things such as battle/siege schedules (depending on game design).
Both Wurm Online and Albion Online have ganking protection: increasing the resistances of a player based on the number of enemies targeting that player. I think this only applies to direct fire, not AoE. I haven't seen it make a huge difference myself, but maybe it helps. I note that it would seem to encourage the use of AoE.
Wurm Online also has spell spam resistance (because spells are relatively powerful in WO): if you get hit with a fireball, you get a temporary fire resistance buff. This does diminish the power of AoE.
The point of schedules is to prevent one elite team from jumping from battle to battle. I'm not sure, however, how much this helps prevent large alliances, since a large alliance can theoretically field more teams.
Now we've talked a lot about friendly fire and collisions as general PvP concepts, especially regarding griefing, but I wanted to revisit them simply in the context of diminishing the power of alliances.
Friendly fire appears primarily to be a way to lessen the utility of AoE. Does that, however, really increase the relative power of a hundred players facing a thousand? The best case scenario is the outnumbered force using AoE freely, and the attacking force unable to do so without a heavy cost.
But, the counters seem obvious: anti-clumping discipline, dumping AoE as the smaller force charges close enough to use their own AoE, and even running into the smaller enemy force to prevent them from using their AoEs without cost. Throw collisions into the mix, and you could also form ranks that couldn't be breached.
So, if friendly fire doesn't help much against lopsided odds, that means it's not much of a discouragement to forming a dominate alliance.
As to collisions, this seems far more useful to the thousand-player side in our hypothetical battle, because they'll be able to block the smaller enemy force against terrain (and keep them from infiltrating their ranks). Something like sneak probably wouldn't help in that case, and would be pretty unrealistic too, I might add.
Again, I'm not seeing how that discourages or hinders a massive alliance.
One more thing that's occurred to me: AoE is all about one player hurting many. Conceptually, this a tactical mechanic that should narrow the power gap between a smaller army and a larger, but the problem, of course, is that both sides have AoE.
I wonder if it would help to decrease the range of AoE attacks (or all ranged attacks) for a cooldown period if you've been moving. This could give the defender a realistic tactical advantage when being assaulted, since they would able to fire AoE at advancing troops before the enemy could return fire. -That- could narrow the power gap.
And in general, I think defenders don't get enough advantages, strategic or tactical, in these games.
War of the Gods
OK, here's the exotic solution I mentioned. Let the player have unrestricted alliances, but if they abuse them, declare a War of the Gods on that alliance, in which terrifying avatars leading hordes of summoned creatures lay waste to the alliance's lands.
Perhaps the developers themselves could even control the divine avatars!
One downside, however, is that that sounds cool enough for an entire planet someday to goad the developers into it. 
That is not a - good - solution.
