The route data-size has nothing to do with how AI are driving. Driving only 34 baseboards is indicative of missing track marks and signals more than anything else.
I say this from experience because I have a route that's well close to 190 miles long end-to-end and my AI drivers will travel most of it before they have AI issues unrelated to the route length with some of them being more likely self-inflicted errors such as sending the drivers down the wrong track or to the wrong station platform.
Spacing and the kind of signals you use are critical to the operation. Using Jointed Rail's SL-series signals as an example, I place Type 05 permissive/advance signals within my longer stretches of track to allow more than one driver to share a track block and to warn drivers of upcoming red lights. Type 04s are used when going from double track to single, and Type 06s are used from going from single to double and face towards the double track.
Use track marks to guide the AI. Drive via and Drive to appear to work better than the Navigate series now, before this was the opposite. Set up track marks before mid-way and at the end of yards to guide AI away from the yards, and use them to guide the AI in terminals where they will do the dumb AI driver things.
Set up your schedule using the Schedule Library and use the related Copy Commands from... to place the schedules in your driver command queue. This will allow you to not only assign routes quickly to drivers, but also troubleshoot driver issues. When placing the Copy Command, always use Insert from schedule, and always use a Wait... 20 seconds to allow the commands to load. Using Wait... 20 seconds when starting also allows the simulation to load up prior to the drivers starting their routes. I found from experience that this helped many, many years ago even in the early days, besides, this lets the scenery and track to load up before driving and everything looks better.
I also use spinning rust and use Seagate Iron Wolf 8-TB Enterprise quality drives. With their large cache, they are quite fast for spinning rust and by keeping them defragmented, Trainz and other programs load up quickly. I had looked at some SSDs, but they are well beyond my budget thanks to the tariffs and now the AI greed going on right now scoffing up any silicon they can get their hands on.
