In other words if both sides see a collision then they collided. if only one side sees it then no collision occurred.
It's (almost?) impossible for them to both see the same collision.
Changing the modeling to reflect what you suggest would be the same as just getting rid of the collision model entirely.
As an example, a few years back a squadie and I were trying to land a fighter on top of a bomber while in flight in the TA or DA. The bomber was just in auto-level, at a stable speed. I'd fly up behind him and above him and match his speed, and then slowly settle down onto his back. I was trying to hook my landing gear in front of his wing roots. I figured that when my prop hit him it would bend and stop and I'd plunk right down into place.
In reality though, while I saw myself "hovering"
on top of him, he still saw me "hovering"
behind his tail. When my wheels touched him (in my view), I'd break up and fall away. From his perspective though, I'd break up and fall away while still behind him; he never saw me touch his plane, and he never took any damage.
We switched positions, and had the exact same result. It looked like he was hovering in position behind me, and then he'd just break up and fall away. I even saw the "smoke" come off his tires like it would if he'd touched down on the runway, even though from my perspective he never touched me (and I took no damage as a result).
The distance we were offset was the result of our combined lag, and since lag is time-related our offset distance was at it's minimum possible (we were flying at a matched, slow, speed, and at the same heading). We were offset about 50yds or less. Had we been flying high speed in opposite headings (head-on) our lag-related separation would have been maximized (he'd have seen himself
more than 50yds from where I was seeing him). Had we both been on high-speed connections our separation would have been less, but still not perfect (I was on 28K dial-up).
Essentially, we were doing our very best to make both of our planes "collide" in a controlled fashion, and found it impossible... Only the guy in back took damage. Had I zoomed under him with my fighter, and pulled up in front of him, I could have missed him on my end and caused him to die of a collision. I could have flown up in front of his nose on my screen, while flying through his tail on his screen.
Because everyone has different lag, and because lag varies at times, it would be (almost?) impossible to learn (with any consistency)to effectively "ram" someone so that both players saw the collision. If your opponent intentionally tries to ram you, you're pretty dang safe. It's when he gets real close and then tries to avoid you where you'd have a higher likelihood of suffering a collision.