I, being mainly a audio programmer, can agree with this whole heartedly. There a number of resources out there but in general it's mostly the transition work that it needed to be completed by the composer. This would be for the most fluid transition anyway.
The programmer could code a algorithm that will fade one song into another but that could get off beat or they could pitch it up and distort it then play a new track (kinda simulating a record skipping).
In general the best way to get results is to have the composer know what system they will be working with and what limitations and advantages it possesses. If it automates transitions then spend more time on the sounds themselves. Basically knowing what parts of the job the audio manager will automate will greatly help with time management. Also knowing the basics behind how it works will help in implementation of said audio. eg do I need a 1second blank space at the ed of my songs so that my transition matrix won't distort the sounds.