Additionally, you don't actually need the optimum, you just need a "good enough" method so that kids aren't on the bus till 10

> It's 165 separate problems, of getting (presumably, on average) 1k students to 1 school

If you have 165 * (1,000 / capacity) busses and drivers, sure.

The problem is that the district in question has for a lot of reasons, many with good intentions, built this as a 100k separate problems. They wanted to abstract student assignment outside of geography and for the most part they have. They've lost the ability to just "push pin" the map because any way you slice it, it's now too much of a tangled spiderweb, because they don't want the map to make sense. They intentionally wanted to break map clusters and patterns. They can't rely on the map because they've built it so that they can't rely on the map. They've obviously built something that they can't scale easily, and it is a travelling salesman nightmare problem in so many of the harder ways.

(They did this to themselves of course, but suggesting that they just start over with push pins on a large enough scale map ignores why they are in the problem in the first place. The map doesn't make sense. There's no way to intuit a 90% solution because there aren't patterns on the map. They didn't want patterns on the map.)