They had people manually pushing it back up the last two humps, and then they had a rope they hooked to the cart to pull it up the last hump.

Source: was one of the people pushing it back over the humps. Surprisingly a sketchy job - you have to really go for it to get it over the middle hump, and if you don't make it then you have to bail really quickly.

And why not just rope?

You'd either need to thread the rope through some kind of guides that would make it follow the contour of the track (which would almost certainly introduce vast friction losses in the force on the rope), or else the force on your rope if you just let it go freely through the air is going to be at vastly the wrong vector for efficient pulling of the sled as it goes over the secondary rises. I suspect you'd find that it becomes unfeasibly hard to pull the sled over the second rises via a rope that is pulled from the top of the primary tower.

It doesn't have enough momentum to fully go up the last 'slope' so it rolls back down with the momentum to get back to the starting position... I believe that's what i gathered from the end of the page.

At least it might have if that pesky physics hadn't gotten in the way, again.

They probably just pull it up with a rope or something.

They just reverse the arrow of entropy real quick to get it back to its starting direction.

Remember to reengage physics before starting the ride with the next rider.