Agreed, but this belies your earlier comment about it being about fairness.
It's not about fairness (except that monsters, all things being equal, have better initiative checks than most non-striker PCs). It's about interest and variety, as well as overall lethality.
Well, it is somewhat about fairness because the monster's inits do bunch up at the beginning of the encounter more if the DM rolls for each group as opposed to rolling for each monster.
If I roll for each monster, they will almost never totally bunch up together as per S'mon's example. Hence, I won't totally wipe out a Defender in the front row (or any other PC) before s/he even gets to act.
I find it a bit unfair to knock unconscious, or otherwise incapacitate a PC without the player getting to at least try to react to it or at least get one round of actions in. This is one of the reasons that petrification in 4E typically occurs after the second failed saving throw.
If it happens, it happens. Stun is a prime example. But, I don't prefer to use rules that might have a tendency to result in these types of situations.
Sure. I do too, but there's a tradeoff between this and juggling too many separate monster inits.
What do you do for minions?
I give minions each their own initiative. We used to use 3x5 cards for init that we'd rotate through, but for the last year now, we've been using a magnetic board with individual magnetic strips with NPC and PC names on them. The NPC magnetic strips are a different color than the PC strips. It has worked out pretty well if the player controlling that board sits next to the DM. The DM can just glance and see which NPC is coming up next and it's made it pretty darn easy. As for players, there's a little magnetic token next to the current PC or NPC, so it's pretty easy to see (even from the furthest chair down the table) whether the token is anywhere near your PC's name. Minimally, players tend to know the approximate location of the magnetic strip for their PC, even if they cannot read the writing on it from far away. If a player rolled a lousy init, it's easy to tell when the token gets near the bottom of the list.
So, 3 monster inits or 12 monster inits, it doesn't really matter. In 3.5, I had upwards of 25 inits at a table on occasion, so it is doable.
I'm of the school of thought that players shouldn't definitively know that a foe is a minion. It breaks verisimilitude for me. So, each minion having their own init and each minion rolling their own damage allows players to eventually figure it out, but not automatically know 100% which NPCs are minions and which are not. I also bring in "tough minions" that take one hit to bloody and two hits to kill (and even more variable minions that might die in 1, 2, or 3 hits depending).
And before some people start complaining that I'm not playing minions right, too bad. I prefer a game with mysteries, even in combat. Players shouldn't know that a foe is a minion, anymore than they should know that a given foe is a brute, or a leader, or an elite, or a solo, or what the AC of the foes are, or anything else. Some players already metagame the system too much. If players make the monster check roll, they get the info supplied by the PHB rules, but they don't get main or specialty role info. They can deduce all they want and might often be correct, but I don't hand it out.