Dang it. I told myself I would stop thinking about this problem and now I keep thinking about it.
I think I may have the means to a solution.
The North Box is moving at -3M per minute. That means it is moving at 3M per minute closer to our box. That means it must eventually come to and go South of our box. At that point it would be moving at +3M per minute. But the "change is consistent" so it can never be anything but -3M.
That puts a time limit on how far away the North Box can be at the starting configuration. We obviously have a problem with infinity here, because that messes everything up.
But this is in an ocean. Considering that our compass (and stopwatch for that matter, but that gets into more complicated matters of relativity) works the way a compass works on Earth, I will presume that the ocean is on Earth.
That sets limits. We are measuring with a laser. We know how far the North Box is at the start. Due to the curvature of the Earth, we can calculate the maximum distance of the North Box. That also sets the maximum time that the change can be consistent.
The maximum distance (and therefore time) is also regulated by how far away the boxes are from shore. Once they hit shore, they will cease that consistent change of movement.
The minimum distance is 4M off shore. The maximum distance is the lesser of the furthest place on Earth from the shore or the longest distance for a laser to hit a distant box. Well, not actually really the actual middle of the ocean, because our system of boxes will be moving at different velocities in different directions, but somewhere sort of middleish.
The West Box is moving away at +4. The East Box is moving away at +1. In order to maintain the measurement of consistent change for the maximum time before one of those boxes moves beyond the horizon, our box would have to be moving west. More than zero. The maximum depends on our position.
See where I am going? To maintain maximum consistent change before the system breaks, we can determine what direction we are moving and to some degree how fast.
To do this mathematically, we should scrap the problems of using the Earth. The ocean is a circle. For that, I think we would need to set either the size of the circle or the distance that our box can detect another box. I think that with either of those parameters, it should be possible to calculate the other.
It may be possible to calculate the time and distance traveled by our box. I suspect we would get a range of possible values rather than an absolute. An interesting question would be whether given such a scenario any result in a singular rather than a range value.