Is doing a bunch of different OB/DB splits even going to matter to your players though? Especially once the random result of the dice comes into play? Will they actually spot the difference between the berserking full OB orc who rolls poorly and who they roll poorly against, vs the full parry orc who goes down before his first swing because they roll 246 and a 92 on the crit vs the two with the more moderate splits, vs the one who used adrenal strength this round?
IMO, you'll get a better effect by having most doing much the same thing, and then maybe picking one or two each round to behave a little more distinctively, and you'll have the time and clear head necessary to actually be able to draw a bit of attention to those one or two because you're not trying to manage 15 completely different things.
Using different weapons doesn't necessitate different splits, and if some have shields and some don't, you can just use the same split and then add the shield bonus at the end for those that have them. The shield-equipped orcs will thus have consistently better DBs, and that should become evident over the course of any battle that runs long enough. Although things like stun and injuries will have a bigger impact anyway.
As for fleeing, managing fleeing enemies is easy enough (they're fleeing!). Sure, pick (or let morale rolls pick for you) which ones flee, and transfer them from their previous group to the "fleeing" group.
Another option is have variable groups. Say you have orcs 1 through 15. On round one, odd orcs do X, even orcs do Y. On round two, orcs 5, 10 and 15 do X, other orcs do Z. On round 3, odd orcs do Y and even orcs do W. On round 4, orcs 4, 8 and 12 do X, the others do Z.
Plenty of variability, and, if the behaviours are distinctive enough, the players will pick up on it and most likely see much more complexity than actually exists.