I did SCIENCE. Here's what I found, in no particular order:
[*:3bgl6h12] Carts with riders travel a lot further (at least 10x more), even if it was powered before the entity started riding.
[*:3bgl6h12]Blocks touching the outer corners of curves (see diagram below) will slow down carts with entities in them. Blocks touching any other part of the tracks will not make any difference and empty carts are not slowed down nearly as much.
- Blocks that will slow down carts with entities
 - Blocks that won't affect distance traveled
Minecarts loose most of their momentum when they go off tracks, even if there is no block under them. They lose more momentum when they are on a block than when falling in mid air, though. A booster works best when the carts meet with one on a slope, so they are exactly parallel. They work somewhat when the carts are unaligned, and they sometimes do not align properly without enough momentum. Minecarts that are inside one another will move infinitely in one direction when off tracks, though they can only be aligned directionally with tracks (aligning them vertically, horizontally, or diagonally), or possible interaction with entities such as mobs or dropped items, though this may stop them (untested personally).
It is currently unknown whether boosters that travel with a cart off a slope will continue to boost it, but it is unlikely in my opinion. Minecarts will not collide end to end with blocks at the end of the track they are on, and will partially clip through the block.
Curved tracks can be used to make two way boosters with relative ease, though if done improperly could cause a loop for the cart, until one of the curves is changed in position with redstone powering
> What happens if you let it run for too long was not tested.
Be sure to always factor this in when you keep minecarts in a state where it keeps getting reboosted: eventually the minecart WILL reverse it's direction once that magical capped speed variable reaches the max for the boosted minecart, I've tested this track http://i35.tinypic.com/2vs5eds.png with this, and it seems to properly handle the cart reversing direction, the one you showed, http://i36.tinypic.com/68zr4l.png , might have the cart get stuck underneath when this happens, usually depends on where the cart is when I starts reversing at max speed.
The booster thing isn't linear because you probably pushed the cart manually before hand, thus screwing results.
Angled tracks seem to have less friction because they're shorter (length = sqrt2/2).
Though the pic shows the cart over the slope, to test it I put the cart directly on the downhill track. I might have put it into different parts of the track, but I don't think that would account for all that difference. And by the way, I repeated the tests a few time just to be sure.
And the insight about angled tracks does make sense.
You travel significantly faster on diagonal tracks than on straight tracks. I haven't tested it extensively, but it looks like a diagonal cart has the same horizontal velocity as a straight cart despite having a longer distance to travel, if that makes any sense.
Let's put it this way: Getting from point A to point B, where A and B are situated diagonally to one another will take twice as long if you take the L-shaped path compared to the diagonal, rather than sqrt(2) as long.