Conceivably, you could put multiple carts on one track, but I haven't tested it. I suspect that they would interfere with one another, some how.
There's at least two ways to make sure the carts don't bump into each other.
First, you can cover the drop shaft with a hatch and link it to a cart activated pressure plate. While cart(1) is traveling through the circuit, cart(2) sits on the closed hatch. When cart(1) reaches the pressure plate, the hatch drops cart(2) into the circuit and closes. Cart(1) comes to a rest on top of the closed hatch, cart(2) triggers the pressure plate, cart(1) drops in, cart(2) stops on top of the hatch and so on.
The downside to this system is that the pressure plate has to be at least 100 ticks of cart travel time away from the hatch so that it has time to close between triggering and the arrival of the next cart. It's also not very fast, feeding one minecart every 100+ ticks at best. I would only use it for very long distance hauling, where you have space for more than one plate-operated hatch stop.
The other, slightly harder way is to drop multiple minecarts on the same square from z+2, where z+0 is the level with the magma reservoir. First cart drops on the submerged impulse ramp (where it must be held by a door or something for the loading process), the rest stack on z+1. When the first is released, it rolls forward and gives space for the next cart to drop in, one at a time. An unsupported minecart seems to take 8 ticks to drop. As long as you make sure the carts keep moving at a steady pace with no decelerations on the way there shouldn't be any collisions. This is the hard part. It took me some time to adjust the distribution of impulse ramps but it's doable. The examples I've read about used powered rollers, which probably makes it easier to pace the carts if you're willing to mess with rollers.
The advantages of this system are that it's obviously much faster, putting out one cart every 8 ticks. There's a slight problem with integrating this into the magma loading system proposed here since cart(1) would collide with cart(2) when it was trying to travel backwards in the reservoir. I suppose it's possible to solve that by making a U-turn and a parallel return path instead of a dead end, so that cart(1) can travel back without bumping into cart(2).