Yes, those would be interesting sequences. For this example, the number of points at each iteration is: 6, 13, 288, ??
But due to it's symmetry, some intersections are shared. For 6 random initial points, the number of intersections will vary depending on the position of the points. But the maximum number appears to be: 6, 15, 774, ??
For 5 random initial points, the maximum number appears to be: 5, 5, 5, 5, ....
For 7 random initial points, the maximum number appears to be: 7, 35, ?? ....
It is worth noting that 5 is the lowest number of initial points which will actually create a continuing pattern (just a star within a star within a star, forever). Anything less will fizzle out at the 2nd iteration. 6 is the lowest number which creates a rich pattern.
If we keep points from previous iterations, 5 is lowest number which will create a rich pattern.
The sequence for this one is: 5, 10, 26, 741, ?? (That's for 5 equally-spaced circle points as our initial points. For 5 random points, the sequence is different)
Also worth noting that we are only in 2d here. Would be interesting to explore limit cases as the number of points and number of dimensions -> infinity.
Also, fun speculation: once we apply deep learning / neural networks to deep math, we might be able to accurately "guess" the next iteration without having to perform every calculation.
Attachments:
