Earlier quoted context omitted.
I’m not sure I follow. All of those additional optimizations are pretty simple problems in the OR field.
I've worked recently in this exact field (known in OR as "pickup dropoff problem with time windows" or PDPTW). As you say, all these modifications are completely standard (especially ones that minimize the worst-case "suffering" of the most-affected passenger).
Here's [0] a paper where they analyze the mistake of the feasibility check that experts in the field failed to do properly. Here's [1] a paper aggregating all the timing problems that arise and their algorithmic complexity. Some of the timing problems, including the constraint of limiting the time of the passenger in the bus had O(n^2) or O(n^3) feasibility checks. That's slow. Especially slow if combined with integer linear programming or branch and cut algorithms.
If your system instead minimizes the riding time by adding a cost function to a constraint, making it soft, in most cases the cost function is so ill defined that the solution no longer does what you want, can hardly minimize all the constraint to a normal solution, and you get a huge mess.
There's no state of the art solution that models these constraints as soft ones.
These problems being standard does not mean that they are simple.
0: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.927...
1: https://w1.cirrelt.ca/~vidalt/papers/Timing-Problems-Final.p...