I didn't play Ars Magica for long enough for the variant “He hits, I soak” to become clichéd, though my converted high-CON ex-AD&D2 Uruk Hai was very good at that version of the old standard. And it depends what you mean by a pattern — if you take the software engineering meaning, that really means a stereotype, in the same sense that that word used to have in printing, of a ready made format Opening gambits are a source of behaviours that often get thus stereotyped — “Here come the flying feet of Kung Fu!” or “We Starsky & Hutch the door.“ — a reference from when they were on the TV (as opposed to the 2004 movie), not that long before the very early D&Ding days — the fighter knocks down the door while a mage stands ready to Magic Missile what may be revealed.
But I guess the real intent of the topic was on the behaviour that emerges when you actually invoke the game mechanics. At this point the fact that each entity in its turn — and that encompasses even the different turn rates for different entities that is part of the Hero system — is given an action that may be used to attack, and that usually meets some implicit resistance from the target. The underlying game mechanics will always drive you to some variant of the the “I attack, he defends” iteration; even in Over the Edge and Feng Shui, no matter what their respective authors would like to have you believe.
All that those games attempt to do is hide this inconvenient fact under layers of fluff. Such spurious detail has for me the effect of reducing both my immersion (as a distraction) and also the credibility of the action — I would rather hide detail I'm no expert in under the abstraction of the mechanics. Otherwise you get to the ludicrous things like in the Theatrix thread on r.g.f.a about 8 years back where what the acting player described in the belief that it is a strong move was resolved by the adjudicator taking the point of view that it was a weak one, or descriptions of a combat in terms of fencing stances (recalling a somewhat more recent thread on u.g.r or perhaps r.g.f.m). Not only am I not au fait with the technical terms, but I do not find it credible that such technically precise stances would be applicable to wilder melee — especially in the example I recall of a combat with a broo. This style of play also seems distressingly full of examples in which a player asserts what his opponent is doing as much as his own character.
No, either you go to systems that plot moves at the sub-100ms increment with Laban Notation, using complex transition rules to simulate physical limitations on how far you can move, and that you can't move one arm though another, and computing whether you'll fall over given that current stance, and how to recover, and then you have to learn how to fight properly, for real, to use it — something I contemplated over 25 years ago, and then drew well back from — or you're stuck with the abstracted action/re-action iteration.