they're all flawed because they all have a flawed internet connection in common. but one is the least flawed effectively:
1) no collisions is not right, that'd mean some tactical loopholes that are probably enough to disqualify this one outright. What's strange is this has not yet been removed from GV modeling, where it's even bigger a loophole.
2) both planes systematically destroyed would mean making players responsible for some other guy's flying, which is just not right, again. You never see where the other player-driven objects really are and thus can't rely on just that. Unless, if netlag was actually predictable and possible to learn, you did, but that wouldn't be much of a 'simulation', and that's not the case: at the scale where planes are likely to touch, netlag is too large to count on acting one way or another.
3) a system where you are only responsible for what you see is flawed by principle but actually gives the best results, considering most internet connections are good enough.
The only common case that should frankly piss off a player is nearly hovering around an opponent on his FE, and for this opponent to maneuver real close (by design since you are trying to bet everything on your outstalling him right at the edge of his ballistics cone) and ram the player by lag from what was initially a safe distance.
In this case the opponent's a 'guilty' rammer: he was the only one with any maneuvering ability, but walks away while the first guy who let his anchor out is sunk by netlag. That's your gamble as a player. Once you have some experience, it's second nature to know what's risky and what's not, and how much risk you want to take; and it's really not so much a handicap compared to real life lagless conditions to come close enough to making this system worse than the other two.
The case of rabbid warping is not too common, and even less that of a player taking advantage of it. You're quickly asked to fix your connection by witnesses and reported if abusing it.
