Lost dog, no baiting there, just most people don't understand the physics going on.
Now that you have the two separate arms configuration in your head and can get a feel of how vectors work it is just an extension of the same principle.
If you do as I and some others have done and tie a loop instead of the line being tied to either side of the spreader you simply change the vectors again.
Imagine that there is no spreader bar and that you are looking at two anchors like in the monkey link.
First imagine that those yellow dots are free to move around rings you can thread a rope through.
Now make a loop of rope that goes through both of the rings so that it forms a triangle with the load and then pull on the load, you will find that both rings will pull towards each other, and that the angle they adopt will split the difference between the load and the bit of rope horizontally between each ring.
This act now changes the angle or vector that the forces are acting on the anchors.
With the setup in the link, the load angle/vector on the anchors (little yellow dots) is simply the angle that the rope heads off towards the load, and the shorter the ropes to the load gets, makes those angle get closer and closer towards the dangerous infinite force straight line between the two anchors.
With the triangle setup, the bit of rope between the two anchors applies another force to the anchors equal in force to the force that the load is applying, so now these two forces add together to create a new vector angle that splits the difference between the two lines of rope. See diagrams.
The blue lines indicate the vector angle of these combined forces for the triangle setup.
Now remember that the wider this angle is, the greater the load applied to the anchor.
You will see by the diagram that the vector angles are much wider in the triangle setup versus the vee setup.
In the climbing world the triangle setup is known as the 'American death triangle' because of the high forces developed doing horrible things like pulling high rated anchors clean out of the rock or breaking slings etc.
Now any rope friction at the corners will tend to turn the triangle to some extent towards the V configuration and shift the vectors slightly favourably downwards.
So a simple setup suddenly gets very complicated to predict without using load cells to measure it all.
So what I did was consider the worst case possible and assume that there was no rope friction and make sure anything I used exceeded the forces calculated.
Don't like the idea of being dropped into the water from a height.
My choice of 90 degrees was a compromise between sliding goodness and a nod of respect towards the dangerous vector gobbledogs.
Note that had I tied off to each side then the forces would have been much lower, almost a third of what I ended up with.
I would have tied off to either side of my spreader bar, but the knots would get in the way of the harness buckle.
So keep in mind when creating your own custom setups that small changes in configuration may bring surprising results, and not always in your favour. . . . .