Here's how I (simplistically) see this issue:
Obviously, and backed up by all of our "the current 'wants' to get back to the source" explanations, there must be just as much current flowing from the load to the source as there is from the source to the load. Plain physics.
What flows only from the source to the load is power. We use the electrical system to transfer energy from one place to another. Just like with phyical wastes, like friction andaerodynamic drag, not all of the energy gets used by the load, but it all must be generated and delivered.
Instead of making a 50-mile long axle shaft from the waterwheel to the millstone, we convert the waterwheel's physical energy into electricity with a generator, and the electricity back into physical energy at the millstone with a motor. The result is work done.
The 50-mile distance bewteen the waterwheel and the millstone is covered by wires that carry current, whether it's AC or DC. That current must flow in both directions; alternately in AC, and continuoulsy in DC. But, it flows equally in both directions.
A clamp-on ammeter or a CT proves that by requiring separation of the circuit conductors to function. If we enclose all conductors, the net result is zero, unless some of that current bypasses the current sensor, which is how a GFCI detects faults.
The point is that we care about getting power from the source to the load. We generate a voltage, and the resulting current is the vehicle that transfers power from one point to the other. The limit is governed by the physical parameters of the equipment.
Any power that does not get used by the load to do the work, but the system will not fiunction without, is wasted energy. Some of it is simple losses, some of it is magnetizing current, some of it is reactive energy. But, it all has to be supplied and carried by the system.
So, my conclusion is that current flows in both directions, and equally so. Reactive power also flows both ways. If it didn't, it would be usable as real power. The real power only flows from the source to the load, and is used to produce the final desired output.