It’s not the capacity, it’s the cost of wiring it. There are over 140,000 miles of trackage in this country. Fiberoptic costs around $50,000 per mile. So you’re at $7B not including the cost of hookup and monitors. If you want a monitor every mile (an overheated axle will go red hot and fail in about 3 miles), you are looking at an enormous cost for installation and maintenance.
By contrast there are 1.6m rail cars in the country. A simple $100 monitor now costs $100m, and installation is “slap it on the railroad truck (wheels)” seems easier.
Except you need power to run them, which means “hardwire”, and there are stretches where you are dozens of miles from a node so a battery doesn’t work for those kinds of remote areas. By contrast a system mounted in the freight car itself only has to reach the cab, and the cab has its own power to run the satellite system. Or, like the vacuum hoses which power the brakes, the engine could power a system through the entire train. They have to be connected already, how hard would it be to add a dedicated wire for sensors to the hookup?