I agree with this because in most (all?) cases, turning "many vs many once" (ex: 5 vs 5 in 1 long fight) into "many vs one many times" (ex: 5 vs 1 in 5 short fights) would be more efficient. However, if there is a method to encourage the players to not change "many vs many" into "many vs one", then we might see many vs many being used more. I'm just not sure yet on what method exactly to encourage the players to embrace many vs many.
Depends, I guess. No individual system will probably work 100% of the time, but applying little things here and there or combining said things could add to the level of difficulty.
1) Zerg the Healer?: Healer's friends get ****** and start spamming knockdowns/knockbacks/stuns/whatever-nasty-move-they-might-not-normally-do.
2) Crowd Control Everything then Many vs. One: Possible immunities build or mobs more readily come with countermeasures to recover their allies.
3) Defensive Barriers: Like that Mithra spell may have hinted, barriers could possibly be set up that'd prevent players from passing through or taking heavy damage in the process unless they negate them by whatever means. Friendlies could pass freely while squishies hide inside while doing their thing at range and more durable mobs bring on their hurt from melee range.
4) Pack Hierarchy: Perhaps attacking certain mobs in a group would have positive or negative multipliers on enmity. Those handling the boss may find minions bee-lining for them regardless of activities others have on the flunkies, while the boss itself may only care more for its better allies.
5) We're gonna kite you!: Fight probably won't be stationary, the enemy taking advantage of its environment and possibly having movement advantages that would force people to juggle the importance of kills depending on threat to the group and how easily they may be caught. I'd almost expect terrain threats here like rock slides, magma/geyser eruptions, strong wind gusts, and so on.
6) Puzzle mobs: Let's say there's a pack of enemies based on the elemental wheel, but they have attacks that can harm their allies of opposite of affinity for considerable damage or even initiating a vulnerability. The trick would be to get this to happen even though the enemy could be actively trying to prevent it. Though, gimmick fights can easily go a number of ways.
7) Personality preferences: Maybe a certain mob will always, without question, target the person with the lowest max HP. A bull could go ******* if someone is wearing red armor. Maybe an internal server counter shows Gladiators have the most kills on its family and will make it a priority to either run from them or hunt them. Could pretty much pick any "reason" for variable behavior here, but ideally it'd make sense in relation to info given in-game instead of more AV mystery BS.
I'm sure there's more an AI could be tailored around, but just the above would give far more fluidity and difficulty to the system over XI's VE/CE or selective, often overpowerable gimmicks.