Wouldn't the north- or south-component of the vectors to C cancel each other the same way they do when going from A–B vs. B–A? If the vector AB is on the north-south axis, and C is somewhere to the east, then the increase in speed from A–C–B would be offset by the decrease on the B–C–A return trip, giving you no information as to what direction is faster.
You absolutely can. Get two synchronized clocks, and start pulsing light at each tick. Then slowly move one clock away and observe how long it takes for the light pulse to arrive.
You don't even need crazy precision for that. A regular atomic clock would do.
As you "slowly" move the two clocks away from one another, the two clocks will no longer be synchronized due to time dilation, and so you need to account for time dilation, but in order to account for time dilation you need to make an assumption about what the one-way speed of light is.
That's why I said "slowly" so that the effect of the time dilation can be kept negligible. If you insist that you MUST account for it, get an additional clock that you move at a different velocity. It'll provide you another data point for the system of equations that you can solve to get the "c" value.
You can measure velocity precisely without any reference to the speed of light by timing how long (according to the clock itself) it takes for it to traverse a given length.
To make it even more watertight, do the experiment in an inertial reference frame with pre-accelerated clocks. Launch two clocks with arbitrary but different velocities along a race track. At the start of the track you have the pulse source along with the reference clock, you synchronize with it as your moving clocks cross the starting line.
Then you measure how long it takes for each moving clock (according to its measurement!) to cross the finish line. This will provide you enough data to solve the equation for the time dilation.
So in my example, the measurement stick that I'm using to mark the race track, might theoretically have different lengths if traversed from south to north or from north to south.
And I believe that this experiment is still possible to do, although not in the special relativity. You need to traverse the same path in a gravity field and compare the accumulated time dilation. It will depend on the absolute distance travelled.
It's an interesting puzzle, but it looks to be impossible to measure the speed of light in one direction due to relativity and time/length dilation effects.
But you can do measure the difference between light going from point A to point B and light going from point A to point C to point B (and adding point C in the reverse direction as well) and I think you can use that to get reasonable confidence that light has constant speed in a given medium in all directions.
But this is totally relevant for trying to figure out asymmetric delivery delays in packet switched networks... You can't measure the delay directly, and if you use the network to synchronize your clocks, you can't use the clocks to measure the delay either, because most synchronization assumes symmetric delay. You need an externally synchronized clock, like a GPS reference etc.