The idea is actually good, but massively expensive. By my calculations, it would take 183.3 seconds to decelerate at 1g from 4000mph to 0mph. That's a lot, but for just about any problem, your train would be past it long before it became a problem, or the computer could detect the problem and slow the train when it's farther out than the 3 minutes it takes to come to a stop.
Energy isn't too much of a concern actually, yes, it'd require a lot of energy, but a lot of that can be regained through regenerative braking just like hybrid cars and electrified railroads do today.
I hesitate to just throw good ideas like this out onto the internet, but given some of the impracticalities, I don't mind about this one: What would be interesting to do is build a few "cargo" maglevs as roll-on roll-off for rail vehicles so maybe an Acela rolls onto the train in New York from Washington D.C., then rolls off in Plymouth to St. Pancras in London giving people a one-seat ride from Washington D.C. to London, England.