Welp, seems like colliding minecarts doesn't have much FUN potential:
i scienced around a bit and as far as i can tell the rules are:
- if a moving cart bumps into a standing cart, it 'transmits' its impulse to the bumped cart and comes to a stop.
- if two moving carts collide, the first-built cart transmits its _non-zero_ impulses (probably listed separately for the three dimensions) to the 'younger' cart and stops where it is.
points in case:
two diagonally-moving carts, one going NE, one SE. If the 'younger' cart is the one going NE, it will move away straight SE, eastward and southward speeds appear to be exactly those of the 'older' cart; its own northward impulse is lost without a trace. Swapping the carts sees the younger one (coming in on a SE heading) going away towards the NE.
Two carts started simultaneously (dropped through hatches onto ramps) and accelerated to a bit over 200k by 30ish ramps collide on a ramp which one is passing going north, the other going east. The 'younger' cart (started going north) flies off NE while the older stops and falls back down the launch ramp. The launched cart's ascent and northward speed are the same as for a north-going cart without the collision, with the collision it just gets an additional eastward impulse. Interestingly, the eastward speed was a fair bit higher than the northward speed (16 tiles north vs. 23 east?), more than the weight difference between chestnut and highwood should be able to explain, and the path and thus speed difference should be minimal...
Taking those same two carts and bashing them against each other in a frontal collision on an E-W track:
the older cart stops dead, the younger cart zips away from the collision site. I swapped the carts around once to make sure it wasn't just path weirdness, and when the 'younger' cart came from the west, it went off to the west, and when it came from the east, it was propelled away to the east as well, so it's certainly 'build order'. In both cases, the impulse the younger cart brought to the meeting got dropped entirely.
Relative cart weight has an influence, but i'll have to go build a different rig to see if a heavy 'push' cart can significantly increase the speed of the pushed cart. With the long ramp i've used, the input speed was too high to really verify possible further speed increases. I could only observe that it worked the other way 'round: a wooden cart will only impart very little speed on a brass cart. Little surprise there.
Edit: well, of course bumping different-weight carts into each other can only reduce, not increase speed: throwing a wooden cart at a fifteen times as heavy brass cart imparts only a small fraction of the wooden cart's speed, bumping a billon cart into a wooden cart with a highest-speed roller only accelerates the wooden cart to max roller speed, no more. Meh.