After thinking a bit, to not say we can't find a workable framework. As I see we can classify targets into 3 categories: Targets, 1st degree targets and 2nd degree targets. I can even make it match my personal understanding of what is warped and what isn't.
Targets: Individuals, Rooms, Groups, Boundaries, etc. Eg. A Pilum of Fire, a person targeted by Wound that Weeps, a sword upon which Edge of the Razor is cast, things protected by an Individual ward, the actual boundary of a boundary spell.
1st degree targets: targets directly derived from Targets and affected by the spell, or things on which the Target relies for it's existence. Eg. People in a group, the individual to which a Part belongs, a sword upon which Blade of Virulent Flame is cast, things inside a Circle.
Targets and 1st degree targets suffer warping.
2nd degree targets: things wich are only tangencially affected by the spell. Eg. A person targeted by a Pilum, people attacked by swords enchanted with Edge of the Razor or Blade of Virulent Flame, things moving through a Hermes Portal, a thing bitten by a person transformed into a wolf, something gazed upon by someone with an Intellego Vision spell, things against which wards ward, things hit by an animated tree.
2nd degree targets do not suffer warping.
2nd degree targets are what I would understand as "colateral targets". Things that are only incidentally affected.
My definitons do not match the cases proposed by David on his first post, of course.
I would not defend that penetration to 2nd degree targets does not benefit from the use of ACs either. Again, to me this distinction (some things you benefit from an AC, other things you don't) makes no sense at all.
I will also point something else. Whatever definition we find. Let's assume we find one all-encompassing definition for colateral targets that matches David's first post. This would still require huge errata, in terms of how big of a change this brings to the rules (not how much text, but how this actually impacts the game' framework). It would be needed to fully define colateral targets (we don't even have a formal definition to what constitutes a target in the books). And then another to make penetration different for targets and colaterals. Then we'd need to check for corner cases inside hermetic magic (magical items, sense spells, wards, etc). And then, probably, it would be wise to comb the books and check if this brings unforeseen changes to things outside of hermetic magic (powers, methods, supernatural abilities, hedge traditions, rival traditions, etc).
Even if we can settle for a workable definition, I do not thing there is any actual benefit to that besides pedantic discussions on the inner workings of magic. Unless we are talking about an hypotetical 6th Ed.