It might also be that "longer worst case wait times" risk turning into a different failure mode entirely. Like maybe if the rush hour train is 5 minutes late it's just uncomfortably packed, but if it's 10 minutes late it gets so packed that letting people off at each station now takes 3x longer than it should, the wait time spikes further, and every other train in the system gets stuck behind that one.

That’s a good summary.

This kind of problem is not easy to solve- it’s not the kind of thing that has a closed form solution. You typically use a discrete event simulator to model the outcome of different strategies. It seems clear to me, though, that the answer is going to be a mix of “maximize number of trains” and “space them out nicely” during rush hour. I bet the optimal strategy trades off those two variables based on some probability distribution of “expected number of available trains” over time, such that if the number of likely available trains is lower you space them out more, and so on.

its more than just that though. it's a legit safety issue. the switches are old and have been breaking down at spectacular rates. if a switch breaks when trains on the line are too close too each other there is a much higher chance for a collision due to one of the trains breaking too soon or too late. spacing them out more evenly reduces the chance of that scenario.

As someone with difficulty in crowded places I would have preferred less crowded trains over anything. Many times I had to wait until the evening rush was over before I could get on the 1 train home - the morning rush I had no choice.