The biggest challenge they have when it comes to diesels, or pretty much any train, is getting enough access to do proper recordings. They've been able to get access to a 37 on the GCR, and could do a lot of recordings. This isn't always the case, so for some trains they have to revert to synthesizing sounds (as with the LIRR M7).
Another challenge is getting all sounds correctly implemented in the game. Some locos have their shortfalls: for the class 166 they made interior sounds, but for some reason aren't using them in-game. (Time constraints?)
There is a longstanding issue with concurrent audio channels: if more than 32 sounds are played simultaneously, the game engine (Unreal Engine 4) stops playing the 'oldest' sound and replaces it with the new one. On some locos this can lead to sounds stuttering or suddenly going away.
It is on their huge "to do list", but as it's a system limitation and they have to cater to a wide audience (all sorts of PCs and consoles) it's not as easy as increasing the number from 32 to something higher.