I have seen several threads about how people have suddenly started having issues with AI drivers not reacting to those commands in the same way that they used to. Things like trains going backwards and ignoring track direction markers, and AI drivers being unable to find things that they had been having no trouble finding before the update. At about the same time, there was a flurry of threads about AI drivers reacting oddly to signals.
These things still work, mostly; they just no longer work exactly the way they used to. The so-called "corner cases" where a couple things both being true at the same time cause a different response than before even while things still work as expected when only one or neither of those things are true.
Trying to explain that more understandably:
When Uncommon Thing One and Uncommon Thing Two are both "normal", you still get "normal" behavior.
When either Uncommon Thing One or Uncommon Thing Two is "unusual", but not both of them, you still get "normal" behavior.
When Uncommon Thing One and Uncommon Thing Two both are "unusual" at the same time, the new code handles this in an unexpected way and you get "unexpected" behavior.
This is to be expected whenever anything as complex as Trainz receives a major new feature.
At least it is in principle. In practice, the slightly-different behavior crops up at inconvenient times and everybody starts complaining about it.
If you are wondering why I put this in such generic terms, it is because there is a lot of this happening with different parts of the program, and we do not yet know much of anything specific about them. I just cannot be specific about details that are still poorly understood.