Number of users connected to satellite/utilization is constant for the entire T_W

Number of users is poisson distributed

The penalty/cost of connecting to a new satellite is a fraction of T_a.

This could in principle be a longer time than T_a but that makes the problem harder and able to grow to infinity since we give more penalty than we can allocate, meaning if we switch every time instances the problem becomes weird.

Routing/finding the end node is handled by the constellation

The position of the constellation satellites is known deterministically

Qeueue length = 0 at start of T_w?